Cover - VU Research Portal

VU Research Portal

Benchmarking Operating Room Performance in Dutch University Medical Centers

van Veen-Berkx, E.

2016

document versionPublisher's PDF, also known as Version of record

Link to publication in VU Research Portal

citation for published version (APA)van Veen-Berkx, E. (2016). Benchmarking Operating Room Performance in Dutch University Medical Centers.

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal ?

Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

E-mail address:[email protected]

Download date: 20. Jul. 2022

https://research.vu.nl/en/publications/c9baeea8-93ec-4dea-90be-8ed0280a6fcb

Benchmarking Operating Room Performance

in

Dutch University Medical Centers

Elizabeth van Veen-Berkx

Layout and printing Ridderprint BV, Ridderkerk, the NetherlandsCover Design Panton BV Ontwerpers voor de ZorgCover photography DigiDaan (DigiDaan.nl)

ISBN 978-94-6299-442-3

© 2016 by Elizabeth van Veen-Berkx, the Netherlands

All rights reserved. No part of this publication may be reproduced or used in any form or by any manner without prior written permission of the author.

Printing of this thesis has been financially supported by VU University Medical Center, Vreelandgroep, Coppa Consultancy, Kiwa Carity, TNO, New Compliance, 4Building, Type 2 Solutions, Mediclabel, Congress Company, Logi Label BV, Mölnlycke Health Care, Olympus Nederland BV and Interflow.

VRIJE UNIVERSITEIT

Benchmarking Operating Room Performance in


ACADEMISCH PROEFSCHRIFT

ter verkrijging van de graad Doctor aande Vrije Universiteit Amsterdam,

op gezag van de rector magnificusprof.dr. V. Subramaniam,

in het openbaar te verdedigenten overstaan van de promotiecommissie

van de Faculteit der Geneeskundeop vrijdag 2 december 2016 om 11.45 uur

in de aula van de universiteit,De Boelelaan 1105

door

Elizabeth van Veen-Berkx

geboren te Vlissingen

promotoren: prof.dr. G. Kazemier prof.dr. H.G. Gooszen

TABLE OF CONTENTS

Introducing the Dutch Operating Room Benchmarking Collaborative 7 Aim and Outline of the Thesis

Chapter 1 Benchmarking Operating Room Departments in the Netherlands 11

Descriptive Studies 39

Chapter 2 Enhancement Opportunities in Operating Room Utilization 41

Chapter 3 The Influence of Anesthesia-Controlled Time on Operating Room Scheduling 63

Chapter 4 Effect of Individual Surgeons and Anesthesiologists on 81 Operating Room Time

Interventional Studies 99

Chapter 5 Successful Interventions to Reduce First-Case Tardiness 101

Chapter 6 Scheduling Anesthesia Time Reduces Case Cancellations 121 and Improves Operating Room Workflow in a University Hospital Setting Chapter 7 Preoperative Cross Functional Teams Improve Operating Room Performance 139

Chapter 8 Multidisciplinary Teamwork Improves Use of the Operating Room 153

Chapter 9 Multidisciplinary Teamwork is an Important Issue to Healthcare Professionals 167

Chapter 10 Dedicated Operating Room for Emergency Surgery 187

Reflections 201

Chapter 11 Twelve Years of Operating Room Benchmarking in the Netherlands 203

Chapter 12 Summary and General Conclusions, General Limitations, 227 General Discussion and Future Perspectives for Research, Lessons Learned

Appendices 255

Letters to the Editor / Commentaries 256Samenvatting 260Acknowledgments 268Dankwoord 270Publications and Presentations 272

Introducing

the Dutch Operating

Room Benchmarking

Collaborative

8

9

Aim and Outline of this Thesis

The general aim of this thesis is to find an answer to the question whether a nationwide long-term benchmarking collaborative of the operating room (OR) departments of all eight University Medical Centers (UMCs) in the Netherlands could lead to improvements in overall OR management. For this purpose, several studies, almost all multicenter, were conducted:

one exploratory study combining qualitative and quantitative research methods;

three descriptive studies based on a substantial amount of multicenter data;

and six quasi-experimental studies to determine the effect of specific interventions in different OR processes.

The first chapter (Chapter 1) provides an introduction to the topic of benchmarking and more specifically to the nationwide operating room benchmarking collaborative, a joint initiative of the eight Dutch university medical centers (UMCs). This exploratory study combines qualitative and quantitative methods and presents key findings useful for benchmarking in (university) hospital settings. Chapter 2 provides insight into improvement potential with regard to the utilization of expensive and limited OR time, based on a multicenter study of several performance indicators and their direct as well as indirect relationships.

Scheduling surgical procedures is a complex process. While progress in OR scheduling methodology has been made over the past years, opportunities for improvement in this area of research still remain. Chapter 3 explores the options to enhance the scheduling of specifically anesthesia-controlled time, while previous studies predominantly indulged in the subject of predicting surgeon-controlled time. Chapter 4 investigates the influence of surgeons and anesthesiologists on OR time and the existence of a work rate effect.

To assess a) the success of interventions implemented to reduce first-case tardiness (Chapter 5), b) the effectiveness of the implementation of a new scheduling method for anesthesia-controlled time (Chapter 6), c) the effects of implementing cross-functional OR scheduling teams (Chapters 7, 8 and 9), and d) the policy outcomes of an approach for reserving OR capacity for emergency surgery (Chapter 10), six studies applied conducted a quasi-experimental time-series design.

Finally, Chapter 11 provides a general overview of the Dutch central OR Benchmark database as well as a summary of tangible and intangible results after twelve years of collaborative benchmarking. In Chapter 12 the results of these studies were converged to summarize, draw conclusions, discuss and to suggest some directions for future research in overall OR management.

1Benchmarking Operating Room Departments

in the Netherlands

Evaluation of a Benchmarking Collaborative

between Eight University Medical Centres

Elizabeth van Veen-Berkx, MScDirk F. de Korne, PhD, MScOlivier S. Olivier, BScRoland A. Bal, PhDGeert Kazemier, MD, PhDfor the Dutch Operating Room Benchmarking Collaborative

Benchmarking:AnInternationalJournal.2016.23(5):1171-92

12

1INTRODUCTION

As in many countries in the world, the health care system in the Netherlands has been intensively reformed over the last decade1, 2. The introduction of more competition in health care was one of the most important changes3-6. This increasingly urges hospital administrators and clinicians to deliver transparent, high quality care with strict financial budgets. The focus on performance improvement sparked the interest of healthcare providers to measure their performance and compare themselves with others in order to be enabled to perform more efficiently in their operational processes7, 8. As indicated by Porter and Teisberg9 in their landmark ‘Redefining Health Care’, competition among health care providers should be focussed on value (defined as “health care results per unit of costs”) and supported by widely available outcome data. Obtaining such data, however, requires appropriate management instruments that can disseminate business information and compare the performance of a single provider to others. Benchmarking, defined as “a process of continuous measuring and comparing an organization’s business against others”10, is described as one of the approaches to obtain useful results11-14.

To assess the application of benchmarking in hospitals, De Korne etal.7 have developed a “4P” conceptual framework (Fig. I).

Figure I. The “4P-model” with key conditions for the application of benchmarking7

13

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

The key conditions, based on literature study, are: (1) purposes (learning from others, identifying performance gaps, implementing best practices); (2) performance indicators (SMART indicators, comparable indicator information, reliable data gathering and sharing); (3) participating organization similarities (in structure, process, outcomes; no competition between participants, voluntary and involved participation); and (4) performance management system (cyclical, internal). The model has been validated in international and U.S. domestic benchmark initiatives between eye hospitals7, 15 but has not been applied in other settings.

Therefore we have studied an on-going collaborative benchmarking initiative between the operating room (OR) departments of eight University Medical Centres (UMCs). In the Netherlands, OR departments of all eight UMC’s established a nationwide benchmarking collaborative in 2004. The objective of the benchmark is to compare the utilization of operating room resources and the economic aspects of operating room performance between the UMCs, with the aim to improve this performance. Each UMC provides their surgical case records to a central OR benchmark database. This extensive database, presently comprising more than 1 million surgical case records, is used to calculate key performance indicators related to the utilization of OR capacity. The results from benchmarking - by name of UMC - are only accessible to the participants. However, the database is also used for multicentre research on OR scheduling topics and OR efficiency, and therefore results from benchmarking are published anonymously8, 16-18.

The aim of this study is to assess if the collaborative, long-term approach of the Dutch OR benchmarking initiative leads to benefits in operating room management and to evaluate if the initiative meets the requirements of the 4P-model. Based on the findings we discuss the applicability of the 4P-model and present key findings useful for benchmarking in (university) hospital settings.

LiteratureLiterature identifies several types of benchmarking: internal and external19, 20. While internal benchmarking focuses on performance measurement and comparing within one organisation over time, external benchmarking can be categorised in competitive, functional, generic and collaborative benchmarking. Competitive, functional and generic benchmarking are commonly conducted independently, while the collaborative approach to traditional benchmarking is performed by groups of organisations that work jointly to achieve the same goals. Collaborative benchmarking entails more than merely comparing performance: organisations share their ideas, approaches, process designs and interventions20, 21. This approach implicates the formation of a voluntary network of health care organisations that cooperate in carrying out the benchmarking study and commit to this long-term21, 22.

Although benchmarking was developed for the business industry, it is increasingly being observed in the public sector23-25. However, empirical research on the use and function of

14

1benchmarking in health organizations is still scarce13, 26. Sower etal.27 have described practical benchmarking experiences at the Bronson Methodist Hospital, the Columbus Children’s Hospital, and the North Mississippi Health Services. They concluded that benchmarking could help close the widening gap between hospitals that deliver exemplary patient service and those that provide lower levels of care.

Extensive research exists regarding hospital benchmarking studies using DEA’s (Data Envelopment Analysis). Benchmarking studies applying the method of DEA are predominantly initiated by the government or regulatory offices (e.g. the Dutch Healthcare Authority) and carried out by a separate academic statistical unit. Datasets are obtained from the Ministries of Health, in other words, ‘external data collectors’. DEA is known as a nonparametric mathematical programming approach for determining a best practice of resource usage and service delivery, and is also focussed on explaining variations in cost efficiency due to a hospital’s environment. Cost-efficiency scores measured by the DEA approach are principally used by policy makers who are interested in budget allocation for hospitals13, 26, 28, 29. The character of benchmarks using DEAs, however, is essentially different from the character of benchmarking collaboratives initiated by hospitals themselves and not by a third external party. Although this type of collaborative benchmarking is increasingly used in hospitals, well-described experiences and systematic empirical research are scarce30.

Several studies though have assessed the efficacy of performance reports in stimulating hospital quality improvement. Hibbard et al.31 found in a large study in Wisconsin that disclosure of performance, in private and public reports, resulted in improvement in the clinical area reported upon. Devers etal.32 indicated different mechanisms that drive hospital quality improvement: regulation, professionalism and market forces; benchmarking and reporting performances is thought to be a key strategy for influencing market forces and, to a lesser extent, professionalism. Also in more government driven systems there is evidence for positive effects of performance reporting. Levay and Waks33 analysed national quality registries in Sweden and describe how professional groups are actively engaged in transparency technologies and found them meaningful, despite initial resistance, and continued discontent with specific aspects of the monitoring systems. In the U.K.’s National Health Service, the use of targets and disclosure of performance have resulted in increasing performance due to forces of reputation management34. Hibbard etal.31 argue that the feedback inherent to both public and private reports will be sufficient to stimulate efforts to improve quality, simply because of professional norms around maintaining standards and self-governance. Therefore, benchmarking has been suggested to be applied broader in hospital care. A recent study by Welborn and Bullington35 indicated that of all process improvement techniques available, benchmarking was found to be the most popular and widely used among a group of award winning US health care organisations35.

15

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

Benchmarking has often been approached as a competitive activity resulting in rankings and with a focus on creating competition between participants as driver for improvement. Since benchmarking was initiated in Japan36 in order to improve competitiveness and since Xerox in 1979 discovered benchmarking as an advantageous management instrument19, it is not unexpected that benchmarking initiatives are typically associated with ‘competition’ instead of ‘collaboration’ between organisations. In literature and previous research, as referred to by Wolfram etal.37, it is generally agreed that the motivation behind benchmarking is to improve and to reduce the performance gap compared with the superior competitor38-41. Due to this competitive nature, most benchmarking studies performed by organisations have been conducted individually.

Already in 1994, however, Mosel and Gift21 as well as Gift etal.22 refer to the need for healthcare providers to consider an alternative to the individual method, which is found in the collaborative approach of benchmarking. Wolfram etal.37 clearly contrast the two approaches:

• the collaborative one is characterized by ‘learning with and from others as aim’, ‘partnership as relationship between participants’, ‘a joint action’ and ‘the visual picture is horizontal and visiting (sharing knowledge fromthekitchen)’;

• the competitive one is characterized by ‘superiority or learning to gain position over the other organisation’, ‘a relationship of rivalry’, ‘a unilateral action to gain position on the ladder of success’ and ‘the visual picture is vertical ranking’.

In the Netherlands, health care reform and the introduction of more competition have been a driver for hospitals to compare themselves to others in the challenge to deliver safe, high quality, transparent, accountable and efficient care. Since reforms in the health care system, vertical ranking is increasingly used in order to provide (hospital) performance information and help patients to choose42. However, these rankings generally tend to compare apples and oranges, because they show observed differences in outcome measures between organisations while outcome measures are bounded by methodological difficulties43, 44. Variation in outcome between organisations is subject to patient case mix, differences in measurement (registration reliability), statistical uncertainty (chance), and real differences in quality of care (structure/process)43, 44. For example, a Dutch hospital can score a high rank in one league table yet at the same time score a low rank in another league table. Therefore, ranking can provide inadequate information and the current public reporting can easily be wrongly interpreted by patients43, 45, 46. Van Dishoeck etal.46 even claim that current outcome indicators, used by the Dutch Healthcare Inspectorate, are not suitable for ranking hospitals because of the influence of random variation.

16

1Origin of the Dutch Operating Room Benchmarking CollaborativeIn 2004, the OR departments of all eight UMCs in the Netherlands established a benchmarking collaborative8, 16, 17. This is a joint initiative of the eight Dutch UMCs. Each UMC provides surgical case records extracted directly from the hospital’s self-reported OR data management system to a central OR benchmark database. This central database is used to calculate key performance indicators of the utilization of OR capacity, e.g. first-case tardiness, turnover time and raw utilization. These performance measurements are shared and benchmarked between the UMCs, which enables the identification of areas of improvement by comparing one’s own performance to that of other, similar organizations.

Operating rooms are of paramount importance to a hospital, given the fact that more than 60% of patients admitted to a hospital are treated in the OR47. Efficient use of OR capacity is pivotal since it is considered a high-cost environment and a limited hospital resource48. Due to the aging population and various developments in surgery, demands for OR facilities are likely to increase. Moreover, due to shortages of qualified OR staff, optimal utilization of ORs is an ever increasing challenge. Benchmarking can be applied to identify improvement potential41 and measure the effectiveness of interventions in an OR environment17.

The initiators knowingly decided to develop a ‘self-led’ collaboration with its own budget (paid for by the eight UMCs themselves) and management, independently from external consultancy organisations and external funding. Independence from external companies as well as external financial resources allows the collaborative to make its own decisions regarding the choice of performance indicators and builds more trust concerning knowledge sharing in a safe learning environment.

The collaborative consists of an organisation containing of a steering committee (head of department of Surgery or Anesthesiology) and a project committee (OR managers) in which all eight UMCs are represented. One full time project manager is hired for planning, organising, securing and managing resources. This project management position is financed by the subscription that is annually paid by the UMCs. The project manager cooperates directly with the members of the project and steering committee on a frequent basis.

METHODS

A mixed-methods design was applied49-51. Based on a literature study, the 4P-model (purposes, performance indicators, participating organisations, performance management system7 was used to evaluate the collaborative in a case study of the OR departments of all eight UMCs in the Netherlands.

We analysed the OR performance data using SPSS Statistics version 21. Data were abstracted from the central OR Benchmark database. Regarding the OR performance

17

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

indicators, all elective, inpatient surgical cases were included. If an OR complex of a single UMC was divided into a main location and sub locations such as a Cancer Centre, Children’s Hospital and Thorax Centre, merely the main (largest) inpatient OR location was included.

The interquartile range (IQR) is a measure of statistical dispersion, which contains the middle 50% of the data (the top and bottom 25% of data are left out)52. It is calculated as the difference between the upper and lower quartiles: IQR = Q3 – Q1.

We conducted ten semi-structured face-to-face interviews with key stakeholders from operating room management (n=3), surgical planning (n=2), operating room nursing (n=1), data management (n=2), policy consultant (n=1) and CEO (n=1). Those key stakeholders originated from five (out of eight) UMCs that represented different parts of the country. The interviews were a maximum of 1.5 hours, transcribed verbatim, and subsequently analysed. The semi-structured face-to-face interviews were guided by a topic list based on the 4P-model. Questions involved purposes of benchmarking, the performance indicators, the reliability in data gathering and sharing, participating organizations and their characteristics and environment, the involvement of participants, the performance management system and the cyclical plan-do-study-act improvement approach.

We performed document analyses to reveal information from management reports, policy documents and performance indicator reports. We analysed the minutes of 40 benchmarking meetings with representatives from all eight hospitals involved and performed observations during two benchmarking focus group study meetings. The interview data, documents and transcripts were analysed by using labels from the 4P-model as well as open labels. The labels were used to code and categorise the transcripts and identify recurrent themes, relying on the theoretical proposition from the 4P-model (as suggested by Yin51). All interviews and observations were conducted by the third author. The preliminary comparative analyses were done separately by the second and third author. To increase the construct validity, the first, fourth and last author reviewed the drafts of the analyses. Data triangulation was used when comparing data gathered from different sources49.

Data triangulation applies multiple sources of information and data to investigate complex situations and to increase the validity of the study. This also implies comparing different findings continuously with findings from other sources in order to contribute to a broader and deeper description and understanding of the case49. In this study there were four empirical sources: first, the quantitative operating room data from the central OR Benchmark database; second, the transcripts of the conducted interviews; third, a review of relevant documents and minutes of meetings; and fourth, the field notes of the observations performed during benchmarking focus group study meetings.

18

1FINDINGS

Assessment of the four key conditionsPurposes. In accordance with the conditions found in the literature, all respondents (n=10) had high expectations and indicated that the benchmark collaborative was focused on learning, sharing knowledge and improving of performance. The third purpose of the 4P-model ‘implementation of best practices’ was mentioned in documents, however, not literally mentioned by respondents. Respondents focused more on improving performance; one way to achieve improvement could be the implementation of best practices. The term ‘performance’ was used by respondents as a collective noun for ‘efficiency’, ‘productivity’, ‘patient safety’, ‘patient satisfaction’ and/or ‘quality of care’.

The partnership, signed by the chairmen of the board of every UMC (Picture I), described three purposes of benchmarking at the start of the collaborative:

1. to compare the utilization of operating room resources as well as the economic aspects of operating room performance and learn from similar organisations, with the aim to improve this performance, as indicated by this respondent:

“Ibelieveitisimportanttocompareyourownperformancewithotherorganisations.EspeciallyinGermany,hospitalchainsshareORdataandareabletoimprovetheirperformance.” (UMC8 manager);

2. to avoid comparing apples to oranges, information and knowledge about the underlying organizational characteristics (see Table 1) and methods/processes is therefore also gathered and shared;

3. to learn about the application of benchmarking in university hospitals.

Additionally, two respondents mentioned:“Itisalwaysinspiringtohavetheopportunitytohavealookinsomeoneelse’sbackyard.” (UMC8 manager and UMC2 OR scheduler)

19

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

Picture I. Partnership agreement signed by the chairmen of the board of every UMC and the University of Twente during the initiation phase of the OR Benchmark collaborative

20

1UMC1 UMC2 UMC3 UMC4 UMC5 UMC6 UMC7 UMC8

Total number of annual surgical cases performed

Inpatient cases

98,859 95,716 132,404 102,789 76,563 73,495 148,209 84,664

1 (elective as well as non-elective) Outpatient cases

27,716 47,357 50,381 23,103 29,470 23,459 62,113

Cancer Centre (with seperate OR location)

17,852

Children’s Hospital (with seperate OR location)

53,458 50,618

2 Elective or non-elective/emergency cases performed (%) included are all inpatient surgical cases excluded are cases performed at seperate OR locations

72/28 69/31 71/29 73/27 77/23 80/20 83/17 71/29

N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD

3 Anesthesia-controlled time (in minutes): anesthesia-controlled time is defined as the sum of the time starting when the patient enters the OR to the time when positioning or skin preparation can begin plus the time starting when the surgical dressing is completed and ending when the patient leaves the OR, in other words, ACT is the sum of anesthesia induction time plus anesthesia emergence time. included are all elective, inpatient surgical cases excluded are cases performed at seperate OR locations

68,213 33 21 67,550 42 25 95,492 44 26 77,125 31 19 60,574 36 21 48,485 33 22 122,625 30 21 65,957 31 19

4 Surgical-controlled time (in minutes): the time starting when patient positioning and/or skin preparation can begin to when surgical dressing is completed.included are all elective, inpatient surgical casesexcluded are cases performed at seperate OR locations

68,213 124 102 67,550 136 112 95,492 131 110 77,125 112 108 60,574 124 107 48,485 131 125 122,625 95 92 65,957 136 109

5 Total case duration (in minutes): patient in to patient out of the OR room. In other words, anesthesia-controlled time plus surgeon-controlled time.included are all elective, inpatient surgical casesexcluded are cases performed at seperate OR locations

68,213 159 111 67,550 178 127 95,492 176 124 77,125 143 120 60,574 161 120 48,485 164 140 122,625 124 105 65,957 166 119

Table 1. Organisational characteristics of the participating UMCs

* UMC1 excluding data year 2010, UMC5 excluding data Thoracic Centre with seperate OR location UMC6 excluding data years 2010-2012

21

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

UMC1 UMC2 UMC3 UMC4 UMC5 UMC6 UMC7 UMC8

Total number of annual surgical cases performed

Inpatient cases

98,859 95,716 132,404 102,789 76,563 73,495 148,209 84,664

1 (elective as well as non-elective) Outpatient cases

27,716 47,357 50,381 23,103 29,470 23,459 62,113

Cancer Centre (with seperate OR location)

17,852

Children’s Hospital (with seperate OR location)

53,458 50,618

2 Elective or non-elective/emergency cases performed (%) included are all inpatient surgical cases excluded are cases performed at seperate OR locations

72/28 69/31 71/29 73/27 77/23 80/20 83/17 71/29

N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD

3 Anesthesia-controlled time (in minutes): anesthesia-controlled time is defined as the sum of the time starting when the patient enters the OR to the time when positioning or skin preparation can begin plus the time starting when the surgical dressing is completed and ending when the patient leaves the OR, in other words, ACT is the sum of anesthesia induction time plus anesthesia emergence time. included are all elective, inpatient surgical cases excluded are cases performed at seperate OR locations

68,213 33 21 67,550 42 25 95,492 44 26 77,125 31 19 60,574 36 21 48,485 33 22 122,625 30 21 65,957 31 19

4 Surgical-controlled time (in minutes): the time starting when patient positioning and/or skin preparation can begin to when surgical dressing is completed.included are all elective, inpatient surgical casesexcluded are cases performed at seperate OR locations

68,213 124 102 67,550 136 112 95,492 131 110 77,125 112 108 60,574 124 107 48,485 131 125 122,625 95 92 65,957 136 109

5 Total case duration (in minutes): patient in to patient out of the OR room. In other words, anesthesia-controlled time plus surgeon-controlled time.included are all elective, inpatient surgical casesexcluded are cases performed at seperate OR locations

68,213 159 111 67,550 178 127 95,492 176 124 77,125 143 120 60,574 161 120 48,485 164 140 122,625 124 105 65,957 166 119

22

1Interestingly, all ten respondents mentioned the purpose of networking. This network aspect in relation with benchmarking was not mentioned as one of the purposes found in literature and incorporated in the 4P-model. Respondents mentioned the national annual conference and the two-monthly focus group study meetings arranged by the project manager as opportunities to network with colleagues from other hospitals, working in the same professional field and dealing with the same professional issues. Afterwards, these networking events were found to make it easier for participants to contact an individual professional working in another hospital, to discuss today’s challenges (the ‘mutual support function’), share more knowledge and organize site visits to each other’s OR departments. This increased the understanding and the learning between members of the network.

Performance indicators. Benchmarking requires SMART indicators (specific, measurable, acceptable, relevant and time-framed), comparable indicator information and reliable data gathering and sharing7, 15. According to the document study, a considerable amount of time and effort was undertaken by the steering committee to develop a partnership agreement during the initiation phase of the collaborative. This agreement creates the foundation for trust and confidentiality between the eight participating hospitals. It describes goals and objectives, requirements, opportunities, organisational structure, finance and possible termination of the partnership. Confidentiality and ownership of benchmarking data are two delicate and important parts of the agreement.

During the first years the collaborative was directly and full time assisted by an independent academic department (University of Twente), in order to develop and harmonise data definitions of OR time periods, uniform methods of data registration and definitions of performance indicators among all participants.

Longitudinal data collection within the OR benchmarking collaboration started in 2005 and is still performed today. Every UMC registers details of each surgical case and time periods – e.g. ‘time patient enters the OR’, ‘time surgery starts’ – since multiple years. These time periods are prospectively and continuously measured, and registered electronically by the nursing staff in each Hospital Information System and validated by the responsible surgeon and anesthesiologist. Each UMC quarterly provides records for all performed surgical cases to a central OR benchmark database. This data focuses on the operating room process, and not on outcomes (e.g. mortality, morbidity) or patient safety (e.g. surgical site infections).

An independent data management centre administers the central OR Benchmark database. This centre provides professional expertise to facilitate the collection and processing of data records, as well as data reliability checks. The centre calculates all key performance indicators, based on the data provided by the UMCs: actual time periods are combined with the total amount of allocated OR session time.

The performance of one OR day, which is generally equal to eight hours of block time

23

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

allocated to a specific surgical department, is commonly evaluated by the indicator ‘raw utilization’. The time when there is no patient present in the OR, so-called ‘non-operative time’, can be evaluated by three performance indicators: first-case tardiness, turnover time and empty operating room time at the end of the day, if cases finish earlier than scheduled. If cases run longer than the regularly scheduled hours of allocated block time, this is termed over-utilized time. All these performance indicators were calculated once per OR day. See Figure 2.Table 2 performance indicators measured in the Dutch OR benchmarking collaborative, including definitions.

Figure 2. Indicators to measure the performance of one OR-day

Raw utilization(i.e. all case durations)

First-casetardiness

cumulativeTurnover

timeEmpty OR time Over-utilized time

Utilized OR time

Over-utilized time

Non-operative time

one OR day(in general) eight hour block time allocated to a specific surgical department

24

1 UMC1 UMC2 UMC3 UMC4 UMC5 UMC6 UMC7 UMC8

Indicator and definition N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD

1 Raw utilization (%): the total amount of time surgical patients are present in the OR, divided by the total amount of allocated block time per day x 100%. Block time (generally 8:00 AM - 16:00 PM) was allocated to a specific surgical department. The definition of raw utilization excluded turnover time and over-utilized OR time.

16,550 84 14 35,756 83 15 50,726 85 16 33,692 80 17 30,260 84 15 25,011 76 20 41,885 84 15 33,477 82 15

2 Early start of the first surgical case of the day (in minutes): the difference between the actual room entry time of the first patient on that day (before 8:00h AM) and the scheduled starting time (generally 8:00h AM).

16,550 6 4 35,756 5 4 50,726 7 5 33,692 8 5 30,260 4 4 25,011 8 8 41,885 6 5 33,477 4 4

3 Frequency early start (%): the percentage of the total number of operating rooms that started too early at a random workday.

16,550 0.05 35,756 0.18 50,726 0.33 33,692 0.24 30,260 0.12 25,011 0.17 41,885 0.24 33,477 0.14

4 First-case tardiness (in minutes): or a ‘late start’ of merely the first surgical case of the day. The difference between the scheduled starting time (generally 8:00 AM) and the actual room entry time of the first patient on that day (per operating room). This value was zero if the case entered the OR early or exactly on the scheduled time. The common scheduled starting time was adjusted in case of an intentionally altered starting time. Every minute of tardiness was calculated.

16,550 30 35 35,756 21 37 50,726 26 40 33,692 36 50 30,260 16 36 25,011 32 48 41,885 19 36 33,477 20 31

5 Frequency late start (%): the percentage of the total number of operating rooms that started too late at a random workday.

16,550 0.71 35,756 0.64 50,726 0.54 33,692 0.59 30,260 0.55 25,011 0.64 41,885 0.63 33,477 0.74

6 Turnover time: the cumulative turnover time per OR day. Turnover time was defined as the time-interval between two succeeding cases; the time between one patient leaving the OR and the next patient entering that OR11, also known as cleaning time.

16,550 29 24 35,756 34 26 50,726 34 27 33,692 40 31 30,260 27 21 25,011 39 31 41,885 34 29 33,477 33 23

7 Frequency of turnovers: the absolute number of turnovers per operating room per workday.

16,550 1.79 35,756 1.58 50,726 1.43 33,692 1.81 30,260 1.52 25,011 1.16 41,885 2.3 33,477 1.46 1

8 Empty OR time at the end of the day (also called “under-utilized time at the end of the day): was quantified by the difference between the actual and scheduled (generally 16:00h) room exit time of the last patient of the day, finishing before 16:00h. The common scheduled finishing time was adjusted in case of an intentionally extended finishing time (extended allocated block time, more than the standard of eight hours).

16,550 56 50 35,756 62 51 50,726 61 53 33,692 64 55 30,260 66 55 25,011 88 64 41,885 54 52 33,477 63 53

9 Frequency under-utilized time (%): the percentage of the total number of operating rooms that finished too early at a random workday.

16,550 0.47 35,756 0.56 50,726 0.43 33,692 0.47 30,260 0.45 25,011 0.52 41,885 0.48 33,477 0.56

10 Over-utilized time: was quantified by the difference between the actual and scheduled (generally 16:00h) room exit time of the last patient of the day, finishing after 16:00h. The common scheduled finishing time was adjusted in case of an intentionally extended finishing time (extended allocated block time, more than the standard of eight hours).

16,550 60 52 35,756 53 51 50,726 62 51 33,692 57 51 30,260 57 51 25,011 71 59 41,885 48 46 33,477 55 51

11 Frequency over-utilized time (%): the percentage of the total number of operating rooms that finished too late at a random workday.

16,550 0.48 35,756 0.38 50,726 0.49 33,692 0.41 30,260 0.42 25,011 0.38 41,885 0.49 33,477 0.35

12 Mean scheduling deviation (%): the percentual difference between the realized and expected/scheduled total case duration, divided by the expected total case duration.

21 11 24 22 12 10

13 Absolute scheduling deviation (%): the absolute difference between the realized and expected/scheduled total case duration, divided by the expected total case duration.

35 29 35 36 28 29

14 Number of surgical cases during the night: the absolute number of sugical cases operated on between midnight 00:00h and 07:00h AM.

1,633 2,942 3,946 3,227 1,990 1,158 2,610 1,866

Table 2. OR performance indicators applied for benchmarking

* UMC1 excluding data year 2010 UMC5 excluding data Thoracic Centre with seperate OR location UMC6 excluding data years 2010-2012

25

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

UMC1 UMC2 UMC3 UMC4 UMC5 UMC6 UMC7 UMC8

Indicator and definition N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD N Mean SD

1 Raw utilization (%): the total amount of time surgical patients are present in the OR, divided by the total amount of allocated block time per day x 100%. Block time (generally 8:00 AM - 16:00 PM) was allocated to a specific surgical department. The definition of raw utilization excluded turnover time and over-utilized OR time.

16,550 84 14 35,756 83 15 50,726 85 16 33,692 80 17 30,260 84 15 25,011 76 20 41,885 84 15 33,477 82 15

2 Early start of the first surgical case of the day (in minutes): the difference between the actual room entry time of the first patient on that day (before 8:00h AM) and the scheduled starting time (generally 8:00h AM).

16,550 6 4 35,756 5 4 50,726 7 5 33,692 8 5 30,260 4 4 25,011 8 8 41,885 6 5 33,477 4 4

3 Frequency early start (%): the percentage of the total number of operating rooms that started too early at a random workday.

16,550 0.05 35,756 0.18 50,726 0.33 33,692 0.24 30,260 0.12 25,011 0.17 41,885 0.24 33,477 0.14

4 First-case tardiness (in minutes): or a ‘late start’ of merely the first surgical case of the day. The difference between the scheduled starting time (generally 8:00 AM) and the actual room entry time of the first patient on that day (per operating room). This value was zero if the case entered the OR early or exactly on the scheduled time. The common scheduled starting time was adjusted in case of an intentionally altered starting time. Every minute of tardiness was calculated.

16,550 30 35 35,756 21 37 50,726 26 40 33,692 36 50 30,260 16 36 25,011 32 48 41,885 19 36 33,477 20 31

5 Frequency late start (%): the percentage of the total number of operating rooms that started too late at a random workday.

16,550 0.71 35,756 0.64 50,726 0.54 33,692 0.59 30,260 0.55 25,011 0.64 41,885 0.63 33,477 0.74

6 Turnover time: the cumulative turnover time per OR day. Turnover time was defined as the time-interval between two succeeding cases; the time between one patient leaving the OR and the next patient entering that OR11, also known as cleaning time.

16,550 29 24 35,756 34 26 50,726 34 27 33,692 40 31 30,260 27 21 25,011 39 31 41,885 34 29 33,477 33 23

7 Frequency of turnovers: the absolute number of turnovers per operating room per workday.

16,550 1.79 35,756 1.58 50,726 1.43 33,692 1.81 30,260 1.52 25,011 1.16 41,885 2.3 33,477 1.46 1

8 Empty OR time at the end of the day (also called “under-utilized time at the end of the day): was quantified by the difference between the actual and scheduled (generally 16:00h) room exit time of the last patient of the day, finishing before 16:00h. The common scheduled finishing time was adjusted in case of an intentionally extended finishing time (extended allocated block time, more than the standard of eight hours).

16,550 56 50 35,756 62 51 50,726 61 53 33,692 64 55 30,260 66 55 25,011 88 64 41,885 54 52 33,477 63 53

9 Frequency under-utilized time (%): the percentage of the total number of operating rooms that finished too early at a random workday.

16,550 0.47 35,756 0.56 50,726 0.43 33,692 0.47 30,260 0.45 25,011 0.52 41,885 0.48 33,477 0.56

10 Over-utilized time: was quantified by the difference between the actual and scheduled (generally 16:00h) room exit time of the last patient of the day, finishing after 16:00h. The common scheduled finishing time was adjusted in case of an intentionally extended finishing time (extended allocated block time, more than the standard of eight hours).

16,550 60 52 35,756 53 51 50,726 62 51 33,692 57 51 30,260 57 51 25,011 71 59 41,885 48 46 33,477 55 51

11 Frequency over-utilized time (%): the percentage of the total number of operating rooms that finished too late at a random workday.

16,550 0.48 35,756 0.38 50,726 0.49 33,692 0.41 30,260 0.42 25,011 0.38 41,885 0.49 33,477 0.35

12 Mean scheduling deviation (%): the percentual difference between the realized and expected/scheduled total case duration, divided by the expected total case duration.

21 11 24 22 12 10

13 Absolute scheduling deviation (%): the absolute difference between the realized and expected/scheduled total case duration, divided by the expected total case duration.

35 29 35 36 28 29

14 Number of surgical cases during the night: the absolute number of sugical cases operated on between midnight 00:00h and 07:00h AM.

1,633 2,942 3,946 3,227 1,990 1,158 2,610 1,866

26

1These performance indicators, combined with (trend) analyses and benchmark reports, are shared between UMCs, which enables the identification of areas of improvement by comparing one’s own performance to that of other similar organizations. The respondents indicated that all benchmarking participants can access the central database at any time using a highly secured web based application/reporting tool.

All respondents (n=10) indicated to be satisfied with the current set of performance indicators.

“I believe every performance indicator which is now measured and benchmarked, is useful” (UMC6data-analyst).

“Fromthestartofthiscollaborativewecarefullydiscussedouroperatingroomprocessesandfromtherewedevelopedtheseindicators,andIthinktheyarestillusefultoapplyinORmanagementtoday”(UMC8manager).

“Thecurrentsetofindicatorsisrelevantanduseful”(UMC6managingdirector).

Five respondents expressed their interest in the development and benchmarking of additional performance indicators regarding case cancellations on the day of surgery, productivity of OR personnel as well as OR cost-prices. The indicator ‘case cancellations on the day of surgery’ was considered to be included in the original series of metrics. However, due to difficulties with respect to harmonisation of the definition and registration method of this indicator, it was not incorporated. Recently the steering committee has decided to expand the current series with two new performance indicators in the course of 2014: e.g. labour productivity (worked hours/OR minutes) and cost-prices (OR cost per minute).

Participants. Referring to the literature, there is no competition between participants, participation is voluntary and involved, and participating organisations have sufficient similarities in structure, process and outcomes7, 15. According to the document analysis, hospitals addressed to participate were all University Medical Centres. Non-university major top-clinical hospitals as well as general hospitals were excluded. There were several reasons for this selection of participants, as supported by results of the document study: a small (eight centres) group is able to build trust between participants at short notice, which facilitates collaborative (inter organizational) learning. The eight UMCs are comparable organizations regarding patient case mix – see Table 3 regarding Exceptional Medical Procedures (WBMV) – and their responsibility for tertiary care, clinical research, education and innovation, which enables a fair comparison. Hence, all benchmark participants showed sufficient similarities in structure, process and outcomes. These participants also share the same interest in (current) issues regarding the OR environment.

27

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

Tabl

e 3.

Law

on

Spe

cial

Med

ical

Pro

cedu

res

(Arti

cle

2 W

BM

W) p

erfo

rmed

per

UM

C in

the

OR

in 2

006

Cod

eU

MC

1U

MC

2U

MC

3U

MC

4U

MC

5U

MC

6U

MC

7U

MC

8

D01

0H

eart

oper

atio

ns96

974

698

159

788

979

837

71.

020

D01

1R

epla

cem

ent o

f aor

ta a

scen

dens

5

D20

0K

idne

y tra

nspl

anta

tions

6697

117

120

111

100

56

D20

2K

idne

y-pa

ncre

as tr

ansp

lant

atio

n re

ceiv

er18

D05

0C

ardi

ovas

cula

r ste

nts

2.04

72.

812

2.92

31.

617

3.54

02.

108

2.22

32.

183

D00

2H

eart

trans

plan

tatio

ns19

18

D70

1Li

ver t

rans

plan

tatio

ns44

3220

D70

3P

artia

l tra

nspl

anta

tion

rece

iver

D03

4H

eart-

lung

tran

spla

ntat

ion

267

21

D93

2C

ochl

ear i

mpl

ants

chi

ldre

n6

73

615

1550

15

D93

0C

ochl

ear i

mpl

ants

adu

lts36

1733

1018

4055

27

D02

0A

ICD

-impl

ants

195

305

281

258

452

9718

4

C11

2N

euro

stim

ulat

ors

for p

ain

man

agem

ent

8916

431

372

17

D66

0N

euro

vasc

ulai

r coi

ling

cere

bral

ane

urys

m32

77

D02

2/D

024

Hea

rt ca

thet

her a

blat

ions

254

276

235

349

474

334

D67

0P

lace

men

t ner

vus

vagu

s st

imul

ator

208

38

D67

1R

epla

cem

ent n

ervu

s va

gus

stim

ulat

or

28

1The respondents affirmed that the aim of this OR benchmarking initiative is learning and therefore the relationship between the UMCs is one of collaboration instead of competition. For six of the eight hospitals, the geographical distance is large enough since they are located in different provinces. Also the two UMCs situated in the same city in the Netherlands confirm that their relationship is not competitive, since there has always been collaboration between the two centres, which is intensifying in the near future because of concentration and task distribution of the most complex care, as well as a possible merger of the two centres.

The OR benchmarking collaborative was initiated by a surgeon and a manager working in one of the UMCs. The other seven centres were personally addressed to participate. Participation was not mandated by the government or other third parties but purely voluntary. The number of participants has remained unchanged since the start of the benchmarking collaboration. During OR benchmark meetings every UMC is represented; this was shown by document analysis of multidisciplinary focus-group study meetings, conferences, steering and project committee meetings. The majority of the interview respondents pronounced to be satisfied with the content of OR benchmark meetings.

PerformanceManagementSystem.The 4P-model identified three conditions for the internal performance management systems of organisations participating in benchmarking7, 15:

1. managers must have knowledge about the performance indicators used and outcomes;2. benchmarking findings have to be communicated to stakeholders in the organisation,

to have any effect on performance;3. benchmarking needs to be incorporated in a continuous quality improvement model:

the plan-do-study-act (PDSA) cycle.

Although all participants can access the central OR benchmark database at any time using a highly secured web-based reporting tool, the project manager also provides participants (solicited and unsolicited) with benchmark analyses and -reports. These reports as well as actualities and urgent subjects concerning the OR, set the agenda for benchmarking meetings. The indicators used for benchmarking are indicators prevailing in general OR management (e.g. utilization rates) and largely integrated in the local decision-making process of the participating UMCs. Four participants (n=4) confirmed that benchmark results are habitually included into management reports for the board of directors. However, the incorporation of benchmark results into the local performance management and reporting system was not self-evident in every participating centre:

“To a limited extent.Benchmark results are sometimesusedormentioned inpresentations.Wedonotincludetheactualbenchmarkresultsintoourstandardmanagement report. We should do so more often” (UMC6 data-analyst and UMC5 OR scheduler)

29

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

“Actually,atthismomentIthinkitwouldbeagoodideatoincludeasummaryofthebenchmarkreportinanewsletter” (UMC5 manager).

Document study revealed that since the start, two-monthly multidisciplinary focus-group study meetings are organised to discuss the results of the data analysis and explore processes and practices “behind the data”. These focus-group study meetings are usually visited by approximately 25 to 30 professionals per meeting from all eight UMCs; these professionals represent OR management, anesthesiologists, surgeons, OR nurses, anesthesia nurses and staff advisors. However, healthcare professionals that visit a focus group study meeting are not perpetual delegates since they are not obliged to visit the following meetings. As mentioned earlier, once per year a national invitational conference is organised to provide a broader learning and knowledge sharing platform. The annual conferences are visited by approximately 200 professionals. Through these meetings, the collaborative tries to involve as many stakeholders and employees as possible in the eight participating hospitals. Through promoting dialogue between the participants a learning environment is created.

Recently, the Dutch OR benchmarking collaborative published a study in The American Journal of Surgery17, showing that 43% of all first operations start at least 5 min later than scheduled and 425,612 minutes are lost because of this annually, which has a respectable economic impact. This study also demonstrated that on an overall level of all UMCs, first-case tardiness has decreased since 2005. Moreover, it showed that four centres implemented successful interventions to reduce tardiness. These UMCs showed a stepwise reduction in variation of first-case tardiness, in other words a decrease in IQR during the years, which indicates an organizational learning effect53. The implemented interventions entailed e.g. providing feedbacks directly when ORs started too late, new agreements between OR and intensive care unit departments concerning ‘intensive care unit bed release’ policy, and a shift in responsibilities regarding transport of patients to the OR. One of the UMCs realised a reduction of 27,392 minutes of first-case tardiness in 1 year. Based on $13.29 per regularly scheduled minute of OR time including labour costs, supply costs, indirect costs, anesthesiologist’s fees, and surgeon’s fees, this implied potential savings of $364,040 in that year.

First-case tardiness is merely one of the performance indicators measured and benchmarked in this collaborative. Even though the improvements in tardiness were driven by the Dutch OR benchmark, the ‘cyclic improvement’-approach needs more attention to guarantee similar achievements and secure the long-term existence of this collaborative.

Alternatively, the central OR benchmark database maintained by the collaborative is frequently used for multicentre research on scheduling and efficiency topics. This research proposes recommendations built on extensive data and statistical analyses, concerning the improvement of OR scheduling. Recent research results considering the influence of anesthesia-controlled time was published in the Canadian Journal of Anesthesia16. This

30

1publication appears to be a start towards opening the discussion on this topic in several participating UMCs. Later on this might become a starting point for improvement.

CONCLUSION

This study investigated whether the collaborative, long-term approach of the Dutch OR benchmarking initiative leads to benefits in operating room management and evaluated if the initiative met the requirements of the 4P-model7, 15. Based on the findings we discuss the applicability of the 4P-model and present key findings useful for benchmarking in (university) hospital settings.

The findings of this investigation show that collaborative benchmarking appears to have benefits different from mainly performance improvement and identification of performance gaps. It is interesting to note that, since 2004, the OR benchmarking initiative still endures after already existing for ten years. A key benefit was pointed out in this recent study by all respondents as ‘the purpose of networking’, on top of the purposes recognized in the 4P-model. The networking events organized by the collaborative were found to make it easier for participants to contact and also visit one another. Apparently, such informal contacts were helpful in spreading knowledge, sharing policy documents and initiating improvement. One reason for this is that they could be used to discuss the tacit components of best practices, that are hard to share in more formal communication media. Respondents were satisfied with the content of these meetings and with the exchange of knowledge in an informal manner, the exchange of experiences including sharing best practices as well as discussing worries and today’s challenges in OR management. It enables understanding and learning from each other. These findings corroborate the idea of De Korne et al.7, 15 that participating in benchmarking offers other advantages, such as generating discussions about how to deliver services and increasing the interaction between participants.

This case study showed that this benchmark largely met all key conditions of the 4P-model 7. However, the ‘cyclical plan-do-study-act improvement approach’, which is the third necessary condition with respect to the internal performancemanagementsystem of organisations participating in benchmarking, was not applied in each activity arranged by the Dutch collaborative. Examples of successful application of this approach, e.g. first-case tardiness17, 54, do exist but a ‘continuing improvement cycle’ was not completely incorporated.

31

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

DISCUSSION

The OR benchmarking collaborative saves the eight participating UMCs from reinventing the wheel regarding several issues high on the agenda of operating room departments. De Korne etal.7, 15 has indicated that “taking part in an international benchmarking initiative is in itself seen as a powerful signal to stakeholders that the organisation is actively working on quality improvement”. Although the OR benchmark is a national initiative, the reputation it builds could be another possible explanation for the long-term commitment of the eight centres to the collaborative. At the end of every year there is a clear decision point whether every UMC wishes to continue its participation the upcoming year and is willing to pay the yearly participation fee charged on the OR budget.

During the initiation phase of the benchmark collaborative, a considerable amount of time (two years) and effort was undertaken by the steering committee to develop a collaboration agreement. As described in the findings, this agreement created the foundation for trust and confidentiality between the eight participating partners, because confidentiality and ownership of benchmarking data are two delicate and important parts of the agreement. These first years were also seized by the development and harmonisation of definitions of performance indicators. Common definitions are an essential base for external benchmarking8, 55. The long-term commitment of the eight centres to the OR benchmark collaborative is exceptional, yet might also be necessary to build and maintain trust between the centres, and also be necessary for uniform data registration and harmonisation of indicator definitions.

Benchmarking has often been approached as a competitive activity resulting in rankings and with a focus on creating competition between participants as driver for improvement. This study, however, clearly shows the advantages of a more collaborative approach. An important difference between public reporting and reporting arranged in this Dutch benchmarking collaborative is the fact that the performance as well as rankings are not publicly available elsewhere than to the eight participating UMCs. When information is publicly and freely available, it will be more difficult to build a relation of trust. This is not surprising, since attempts to increase transparency of professional work represent a potential threat to professional autonomy and therefore, professionals often react with suspicion and a certain amount of resistance33. However, when professionals are actively involved in transparency technologies through translation and negotiation in expert networks, public quality reporting can actually become acceptable and advantageous. Advantageous with regard to retaining control over (external) evaluation criteria and drawing attention to professional activities and improvement efforts in order to gain legitimacy and support from external actors33.

From the very first start, the initiators of the Dutch OR benchmarking collaborative as described in this study consistently and literally have avoided ‘naming and shaming’ through publishing and vertical ranking of the eight UMCs, regarding the performance indicators

32

1measured. Lots of attention has been given to honest assessment and avoiding to compare apples and oranges. The physical, organisational characteristics and structure of all participating operating room departments can be very different from one another. Contingency theory claims there is not “one best way for organising” because this is subject to the internal and external conditions of every organisation56-58. Differences in organisational characteristics derive from differences in organisational conditions. Therefore, performance indicators used for benchmarking should take into account these differences, to avoid inaccurate interpretation of observed differences between organisations and to accomplish an honest comparison.

The character of benchmarks using DEAs is essentially different from the character of the Dutch OR Benchmarking Collaborative since it was initiated by the eight university hospitals themselves and not by a third external party. Moreover, data is derived from the local Hospital Information Systems, which are used for daily registration practices. The Dutch OR Benchmarking Collaborative is a ‘self-led’ and voluntary collaboration with its own budget (paid for by the eight hospitals themselves). OR benchmark data is merely used by the participants and not by policy makers, the government or regulatory offices.

Another foundation of the collaborative benchmark described in this study, is the pursuit to learn from the organisational differences in structure, process designs, methods and performance. These differences can be a source of learning as they allow practitioners to compare relations between organisational characteristics and performance, especially in informal settings and networking. These differences also offer every participating OR department the opportunity to engage their own quality improvement pathway. Improvement starts with quantitative analyses and therefore performance indicators should be SMART. In this collaborative the interorganisational or ‘joint learning process’ is more important than ranking participants or to identify ‘the best practice’. The OR departments of the eight UMCs are all providing the same healthcare product: perioperative care in a university hospital setting. It is important to gain insight into managing and controlling this process as well as insight into performance differences, to realise the ‘best fit’ for each OR department.

The Dutch OR benchmark collaborative bears a resemblance to ‘quality improvement collaboratives’ that became popular as the ‘breakthrough series’, an improvement method developed by the Institute for Healthcare Improvement (IHI) in Boston59. Nembhard 60 describes these collaboratives as “structured programmes in which multidisciplinary teams from different organizations work to improve care in one area of their operations (e.g. infection control). As part of a collaborative, teams attend a series of meetings where they learn about best practices in their target area, quality improvement techniques, and the experiences of others that have implemented new practices”. The OR benchmark focus group study meetings as described in our findings share the same goals as the collaboratives and have many similarities.

Nevertheless, the OR benchmark collaborative could learn from the IHI breakthrough series approach to develop a more structured PDSA-approach. Specifically with regard to the

33

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

commitment of the participants during study meetings and the (learning) activities in between physical meetings. When a healthcare professional decides to participate in a breakthrough series, he commits to participate actively throughout the limited collaborative period. In general, this period is limited to 6 - 18 months, which is supposed to drive change. In between physical meetings, teams are expected to implement changes in their own organisation and it is mandatory to share implementation experiences with each other for collective learning through conference calls or digital (internet) platforms60-62. Clearly, this kind of ‘stable commitment’ through continuous participation was not established in the OR benchmark collaborative in this recent study. Healthcare professionals that visit a focus group study meeting are not perpetual delegates since they are not obliged to visit the following meetings. The responsibility for improvement was kept an individual responsibility of each single UMC and not a collaborative responsibility. Future research should therefore concentrate on the investigation of the relation between benchmarking as instrument and the actual performance improvements realised through benchmarking in the local UMC’s.

The current study has the limitations accompanied with any qualitative research and particularly related to interviewing49, 63. Firstly, qualitative research findings must be viewed within the context of the conducted case study49, 63. The perceptions and experiences of the respondents in this university hospital context in the Netherlands might not be transferable to other (general) hospital settings or other countries. Secondly, the number of conducted interviews is restricted; nevertheless, all other data sources are extensive. Thirdly, while all members of the project committee and regular visitors of the OR benchmark meetings were invited to interview, it might be possible that ‘contribution bias’ was present with the respondents who reacted the quickest, being hypothetically those who had more interest in the benchmark collaborative. Despite these limitations, this study provides valuable insights of experiences with benchmarking from a variety of participating centres representing different parts of the Netherlands. The context of the case study and conditions under which this specific benchmarking process took place, was comprehensively outlined, to allow for transferring of results to other settings. In order to increase the validity of the study, data triangulation was applied and a variety of data sources were used. Moreover, construct validity was ensured by deploying several researchers to evaluate the analyses, operating separately from one another49, 63.

Benchmarking is defined as a ‘continuous process’10 and encourages the use of a continuous quality improvement model (the PDSA cycle). Although this OR benchmark initiative, as many benchmark initiatives64, started with a stated aim to improve, actual (measurable) quality or performance improvements are not necessary for this initiative to endure. These findings further support the idea of De Korne etal.7, 15 that benchmarking is relying on iterative and social processes in combination with structured and rational process of performance comparison. The relatively limited focus on OR utilization in this benchmark seems to be

34

1a starting point for exchanging a variety of information and experiences considering the structure, process and performance of OR departments. More attention needs to be given to the relation between benchmarking as instrument and the actual performance improvements realised through benchmarking in the local UMC’s. A collaborative approach in benchmarking can be effective because participants use its knowledge-sharing infrastructure which enables operational, tactical and strategic learning. Organisational learning is to the advantage of overall operating room management. Benchmarking appears to be a useful instrument in enabling hospitals to learn from each other, to initiate performance improvements and catalyse knowledge-sharing.

35

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

REFERENCES

1 . Schut FT, van de Ven WP. Effects of purchaser competition in the Dutch health system: is the glass half full or half empty? Health Econ Policy Law 2011;6:109-23.

2 . Van de Bovenkamp H, De Mul M, Quartz J, Weggelaar-Jansen A, Bal R. Institutional Layering in Governing Healthcare Quality. Public Administration 2013.

3 . Enthoven AC, Tollen LA. Competition in health care: it takes systems to pursue quality and efficiency. Health Aff (Millwood) 2005;Suppl Web Exclusives:W5-420-33.

4 . Scanlon DP, Swaminathan S, Lee W, Chernew M. Does competition improve health care quality? Health Serv Res 2008;43:1931-51.

5 . Schreyogg J, Stargardt T, Tiemann O, Busse R. Methods to determine reimbursement rates for diagnosis related groups (DRG): a comparison of nine European countries. Health Care Manag Sci 2006;9:215-23.

6 . Schut FT, Van de Ven WP. Rationing and competition in the Dutch health-care system. Health Economics 2005;14:S59-74.

7 . De Korne DF, Sol KJ, van Wijngaarden JD, et al. Evaluation of an international benchmarking initiative in nine eye hospitals. Health Care Manage Rev 2010;35:23-35.

8 . Kazemier G, van Veen-Berkx E. Comment on “Identification and use of operating room efficiency indicators: the problem of definition”. Can J Surg 2013;56:E103-4.

9 . Porter ME, Teisberg EO. Redefining health care : creating value-based competition on results. Boston, Mass.: Harvard Business School Press; 2006.

1 0. APQC. Benchmarking Code of Conduct. Guidelines and Ethics for Benchmarkers. 2008.

1 1. Zairi M, Ahmed P. Benchmarking maturity as we approach the next millennium. Total Quality Management Journal 1999;4:810-6.

1 2. Zairi M, Leonard P. Practical benchmarking : the complete guide. London: Chapman & Hall; 1994.

1 3. Blank JL, Valdmanis VG. Productivity in hospital industry. Adv Health Econ Health Serv Res 2008;18:3-12.

1 4. Dattakumar R, Jagadeesh R. A review of literature on benchmarking. Benchmarking 2003;10:176-209.

1 5. De Korne DF, van Wijngaarden JD, Sol KJ, et al. Hospital benchmarking: are U.S. eye hospitals ready? Health Care Manage Rev 2012;37:187-98.

1 6. van Veen-Berkx E, Bitter J, Elkhuizen SG, et al. The influence of anesthesia-controlled time on operating room scheduling in Dutch university medical centres. Can J Anaesth 2014.

1 7. Van Veen-Berkx E, Elkhuizen SG, Kalkman CJ, Buhre WF, Kazemier G. Successful interventions to reduce first-case tardiness in Dutch university medical centers: Results of a nationwide operating room benchmark study. Am J Surg 2014.

1 8. van Veen-Berkx E, Elkhuizen SG, van Logten S, et al. Enhancement opportunities in operating room utilization; with a statistical appendix. J Surg Res 2014.

1 9. Camp RC, Tweet AG. Benchmarking applied to health care. Jt Comm J Qual Improv 1994;20:229-38.

2 0. Commission J. Benchmarking in health care. 2nd ed. Oakbrook Terrace, lL: Joint Commission Resources; 2012.

2 1. Mosel D, Gift B. Collaborative benchmarking in health care. Jt Comm J Qual Improv 1994;20:239-49.

36

12 2. Gift RG, Stoddart TD, Wilson KB. Collaborative benchmarking in a healthcare system. Healthc

Financ Manage 1994;48:80-2, 4-6, 8.

2 3. Booth M, James P, Stevanovic V. Benchmarking of hospital activity data: an international comparison. Benchmarking: An International Journal 2005;12:515-22.

2 4. Guven-Uslu P. Benchmarking in health services. Benchmarking: An International Journal 2005;12:293-309.

2 5. Pantouvakis A, Mpogiatzdis P. Measuring clinical department efficiency: the impact of clinical leadership satisfaction: an application to public hospitals. Benchmarking: An International Journal 2013;20:305-21.

2 6. Blank JL. Innovations and productivity: an empirical investigation in Dutch hospital industry. Adv Health Econ Health Serv Res 2008;18:89-109.

2 7. Sower VE, Duffy JA, Kohers G. Benchmarking for hospitals : achieving best-in-class performance without having to reinvent the wheel. Milwaukee, Wis.: ASQ Quality Press; 2008.

2 8. Blank JL, Valdmanis VG. Environmental factors and productivity on Dutch hospitals: a semi-parametric approach. Health Care Manag Sci 2010;13:27-34.

2 9. Kacak H, Ozcan YA, Lins MPE, Silva ACM, Fiszman R. Determinants of Efficiency for Teaching Hospitals in Brazil. International JOurnal of Healthcare Management 2014;7:60-8.

3 0. van Lent WA, de Beer RD, van Harten WH. International benchmarking of specialty hospitals. A series of case studies on comprehensive cancer centres. BMC Health Serv Res 2010;10:253.

3 1. Hibbard JH, Stockard J, Tusler M. Hospital performance reports: impact on quality, market share, and reputation. Health Aff (Millwood) 2005;24:1150-60.

3 2. Devers KJ, Pham HH, Liu G. What is driving hospitals’ patient-safety efforts? Health Aff (Millwood) 2004;23:103-15.

3 3. Levay C, Waks C. Professions and the Pursuit of Transparency in Healthcare: Two Cases of Soft Autonomy. Organ Stud 2009;30:509-27.

3 4. Bevan G, Hood C. Have targets improved performance in the English NHS? BMJ 2006;332:419-22.

3 5. Welborn C, Bullington K. Benchmarking award winning health care organizations in the USA. Benchmarking: An International Journal 2013;20:765-76.

3 6. Zairi M. Competitive benchmarking : an executive guide. Letchworth, Hertfordshire, England: Technical Communications (Pub.); 1992.

3 7. Wolfram Cox J, Mann L, Samson D. Benchmarking as a Mixed Metaphor: Disentangling Assumptions of Competition and Collaboration. Journal of Management Studies 1997;34:285-314.

3 8. Camp R. Benchmarking: The search for the best practices that lead to superior performance. New York: QR 1989.

3 9. McNair CJ, Leibfried KHJ. Benchmarking : a tool for continuous improvement. 1st ed. New York, NY: HarperBusiness; 1992.

4 0. Spendolini MJ. The benchmark book. New York: Amacom 1992.

4 1. Watson GH. Strategic benchmarking: How to rate your company’s performance against the world’s best. New York: John Wiley & Sons 1993.

4 2. Quartz J, Wallenburg I, Bal R. The performativity of rankings. On the organizational effects of hospital league tables. Rotterdam: iBMG; 2013 Augustus 2013.

37

Ben

chm

arki

ng O

pera

ting

Roo

m D

epar

tmen

ts

4 3. Lingsma HF, Steyerberg EW, Eijkemans MJ, et al. Comparing and ranking hospitals based on outcome: results from The Netherlands Stroke Survey. QJM : monthly journal of the Association of Physicians 2010;103:99-108.

4 4. Mant J. Process versus outcome indicators in the assessment of quality of health care. Int J Qual Health Care 2001;13:475-80.

45. Giard RW. [Top 100 hospitals: varying lists, varying reputations]. Ned Tijdschr Geneeskd 2006;150:2355-8.

46. Van Dishoeck AM, Lingsma HF, Mackenbach JP, Steyerberg EW. Random variation and rankability of hospitals using outcome indicators. BMJ Qual Saf 2011;20:869-74.

47. Eijkemans MJ, van Houdenhoven M, Nguyen T, Boersma E, Steyerberg EW, Kazemier G. Predicting the unpredictable: a new prediction model for operating room times using individual characteristics and the surgeon’s estimate. Anesthesiology 2010;112:41-9.

48. Marjamaa R, Vakkuri A, Kirvela O. Operating room management: why, how and by whom? Acta Anaesthesiol Scand 2008;52:596-600.

49. Bowling A. Research Methods in Health. Investigating Health and Health Services. Berkshire: Open University Press 2009.

50. Creswell JW. Research Design. Qualitative, Quantitative, and Mixed Methods Approaches: SAGE Publications; 2002.

51. Yin RK. Case Study Research: Design and Methods: SAGE Publications; 2009.

52. Field A. Discovering Statistics using IBM SPSS Statistics. Thousand Oaks, CA: SAGE Publications Ltd; 2013.

53. Sehwail L, de Yong C. Six Sigma in Health Care. International Journal of Health Care Quality Assurance 2003;16:1-5.

54. Van Veen-Berkx E, Elkhuizen SG, Kalkman CJ, Buhre WF, Kazemier G. Successful Interventions to Reduce First-Case Tardiness in Dutch University Medical Centers: Results of a Nationwide Operating Room Benchmark Study. Am J Surg 2013;In press.

55. Fixler T, Wright JG. Identification and use of operating room efficiency indicators: the problem of definition. Can J Surg 2013;56:224-6.

56. Burns T, Stalker GM. The management of innovation. London: Tavistock; 1961.

57. Lawrence PR, Lorsch JW. Organization and environment: managing differentation and integration. Homewood (Ill.): Irwin; 1969.

58. Mintzberg H. The structuring of organisations. New York: Prentice-Hall; 1979.

59. paper. IISw. The Breakthrough Series: IHI’s Collaborative Model for Achieving Breakthrough Improvement. Boston: Institute for Healthcare Improvement; 2003.

60. Nembhard IM. All teach, all learn, all improve?: the role of interorganizational learning in quality improvement collaboratives. Health Care Manage Rev 2012;37:154-64.

61. Kilo CM. A framework for collaborative improvement: lessons from the Institute for Healthcare Improvement’s Breakthrough Series. Qual Manag Health Care 1998;6:1-13.

62. Kilo CM. Improving care through collaboration. Pediatrics 1999;103:384-93.

63. Silverman D. Doing Qualitative Research: A Practical Handbook. London: SAGE Publications; 2000.

64. Askim J, Johnsen A, Christophersen KA. Factors behind organizational learning from benchmarking: Experiences from norwegian municipal benchmarking networks. J Publ Adm Res Theor 2008;18:297-320.

Descriptive Studies

2Enhancement Opportunities in

Operating Room Utilization

with a Statistical Appendix

Elizabeth van Veen-Berkx, MSc Sylvia G. Elkhuizen, PhDSanne van Logten, MScWolfgang F. Buhre, MD, PhDCor J. Kalkman, MD, PhDHein G. Gooszen, MD, PhDGeert Kazemier, MD, PhDfor the Dutch Operating Room Benchmarking Collaborative

JournalofSurgicalResearch.2015.194:43-51e1-2.

42

2

ABSTRACT

Background: The purpose of this study was to assess the direct and indirect relationships between first-case tardiness (or ‘late start’), turnover time, underutilized OR time and raw utilization, as well as to determine which indicator had the most negative impact on operating room (OR) utilization to identify improvement potential. Furthermore, we studied the indirect relationships of the three indicators of ‘non-operative’ time on OR utilization, to recognize possible ‘trickle down’ effects during the day.

Materials and methods: (Multiple) linear regression analysis and mediation effect analysis were applied to a dataset from all eight University Medical Centers in the Netherlands. This dataset consisted of 190,071 OR days (on which 623,871 surgical cases were performed).

Results: Underutilized OR time at the end of the day had the strongest influence on raw utilization, followed by late start and turnover time. The relationships between the three ‘non-operative’ time indicators were negligible. The impact of the partial indirect effects of ‘non-operative’ time indicators on raw utilization were statistically significant, but relatively small. The ‘trickle down’ effect that late start can cause resulting in an increased delay as the day progresses, was not supported by our results.

Conclusions: The study findings clearly suggest that OR utilization can be improved by focusing on the reduction of underutilized OR time. Improving the prediction of total procedure time, improving OR scheduling by e.g. altering the sequencing of operations, changing patient cancellation policies, and flexible staffing of ORs adjusted to patient needs are means to reduce ‘non-operative’ time.

43

Enh

ance

men

t Opp

ortu

nitie

s in

Ope

ratin

g R

oom

Util

izat

ion1. INTRODUCTION

Healthcare today is faced with several challenges: rising costs, changing demographics, aging population, technological innovations and changing patients’ demands. Hospitals and operating room (OR) departments in particular, aim to improve quality and safety as well as utilization and efficiency. Operating rooms are cost-intensive, multi-professional parts of health care organizations1. Generally, more than 60% of patients admitted to the hospital are treated in the OR2. ORs typically account for more than 40% of a hospital’s total revenues and a similarly large proportion of its total expenses3. Thus, efficient usage of OR capacity is pivotal.

In ORs, inefficiencies can occur at several moments throughout the day, before, during, between, and after cases4, 5. OR capacity is often evaluated by the indicator ‘raw utilization’, which is the percentage of allocated OR time that a patient was physically present in the room1. The time when there is no patient present in the OR, so-called ‘non-operative’ time, is the sum of three performance indicators: first-case tardiness (or ‘late start’ as it is referred to in the rest of this article), turnover time and underutilized OR time.

Several studies have evaluated OR utilization, mainly by analyzing one aspect of ‘non-operative’ time, such as late start5-10 and turnover time11-13 or the aspects of under- and over-utilized time at the end of the day14, 15. Most of these studies have focused merely on one hospital, a small number of surgical departments or simulation of data. Multicenter studies using an extensive empirical dataset in view of evaluating OR inefficiencies are scarce. Besides, previous studies have not yet evaluated the way in which all performance indicators interact.

We hypothesized that the three indicators of ‘non-operative’ time may each negatively impact OR utilization. Therefore we determined the relationship between late start, turnover time, underutilized time and OR utilization, in all eight University Medical Centers (UMCs) in the Netherlands. We assessed which indicator had the most negative impact on OR utilization to identify improvement potential. Furthermore, we studied the indirect relationships of the three indicators of ‘non-operative’ time on OR utilization, to recognize possible ‘trickle down’ effects during the day.

2. MATERIALS AND METHODS

2.1 Research settingIn 2004, the OR departments of all eight University Medical Centers (UMCs) in the Netherlands established a benchmarking collaborative, which has been active up to today. The objective is to improve OR performance by mutual learning from best practices. Each UMC provides data

44

2

on all surgical cases performed in the individual center to a central OR Benchmark database. This extensive database – today containing more than one million records of surgical cases – is used to calculate key performance indicators of the utilization of OR capacity. These indicators are based on internationally recognized definitions16-18.

2.2 Performance indicatorsOperating room time was evaluated by the indicator ‘raw utilization’ (%), which was defined as the total amount of time patients are present in the OR, divided by the total amount of allocated block time (generally from 8:00h until 16:00h) per day x 100%. Block time was allocated to a specific surgical department. The definition of raw utilization excluded turnover time and over-utilized OR time1, 19. Raw utilization was calculated considering all cases operated on within block time, whether it were elective or emergency cases. However, emergency cases which started after block time were not considered for calculating any of the performance indicators.

‘Non-operative’ time was assessed by three performance indicators: first-case tardiness (or ‘late start’), turnover time and underutilized OR time. The indicator first-case tardiness (a ‘late start’ of merely the first surgical case of the day) was defined by the difference in minutes between the scheduled starting time (generally 8:00 AM) and the actual room entry time of the first patient on that day (per operating room). This value was zero if the case entered the OR early or exactly on the scheduled time6, 19. The common scheduled starting time was adjusted in case of an intentionally altered starting time5.

The indicator turnover time represented the cumulative turnover time in minutes per OR day. Turnover time was defined as the time-interval between two succeeding cases; the time between one patient leaving the OR and the next patient entering that OR11, also known as cleaning time20.

Underutilized OR time at the end of the day was quantified by the difference in minutes between the actual and scheduled (generally 16:00h) room exit time of the last patient of the day, finishing before 16:00h21. The common scheduled finishing time was adjusted in case of an intentionally extended finishing time.

Raw utilization, late start, turnover time and underutilized time are indicators measured once per OR-day, meaning: once per operating room per weekday per hospital (e.g. if an UMC facilitates 20 operating rooms, 20 OR-days were recorded per weekday, if all of these 20 operating rooms were staffed that particular day and allocated to a specific surgical department). One OR-day is generally equal to eight hours of block time allocated to a specific surgical department in a specific operating room. An OR-day was defined as a combination of one operating room and one date on which at least one surgical case was performed. Block-time was not allocated during weekends or holidays, thus performance indicators were only measured during weekdays.

45

Enh

ance

men

t Opp

ortu

nitie

s in

Ope

ratin

g R

oom

Util

izat

ion2.3 Data collection

Data was prospectively collected and analyzed retrospectively for the purpose of this study.All data was registered electronically by the OR nursing staff in the Hospital Information System and validated by the surgeon and anesthesiologist in charge after completion of the operation. Since 2005, anonymized data records of all surgical cases performed at eight UMCs are sent to a central OR Benchmark database5, 22, 23. At the start of the collaborative, data definitions of time intervals were harmonized among all benchmarking participants22, 24.

The collaborative cooperates with an independent data management center, which administers the longitudinal data collection to the central OR benchmark database (Julius Center for Health Sciences, Utrecht, the Netherlands). This center provides professional expertise to facilitate the processing of data records and performs reliability checks before data are ready for analysis. Reliability checks consist of: a check for missing values; a consistency check to determine if data is in accordance with earlier data deliveries; the correctness of data was studied to check if values are outside of a designated range; extreme values were removed from the dataset according to predetermined outlier filtering rules (e.g. OR utilization 25 ≥ x ≤ 110%; late start 0 > x ≤ 240 minutes; turnover time 0 > x ≤ 120 minutes; underutilized time 0 > x ≤ 240 minutes).

Raw utilization, late start, turnover time and underutilized time were sequentially recorded for successive OR days over seven years. The original central OR Benchmark database consisted of a total of 289,977 OR-days on which 986,649 surgical cases were performed at eight UMCs over a seven-years period from 2005 to 2011.

To define a consistent group of data, only in-patient cases were included and all out-patient cases were excluded. In the Netherlands, in contrast to in-patient surgery, the out-patient surgery workflow varies significantly from center to center (different scheduling team, planning horizon and planning methodology). That is why the out-patient OR process is considered as a distinct process, which should be analyzed separately. In some Dutch UMCs, large OR departments are divided into a main (the largest) OR location and different sub OR locations. Sub locations such as a Children’s Hospital, Cancer Center or Thorax Center were also excluded because these sub locations are separate organizational units. OR-days with a missing registration of the specific OR-location and labeled as ‘location unknown’ were excluded. A total of 190,071 OR- days were left for further statistical analyses1.

2.4 Statistical analysis (see also enclosed Statistical Appendix)Data analysis was performed using SPSS Statistics 20 (IBM SPSS Statistics for Windows, version 20.0, IBM Corp. Released 2011.; Armonk, NY, USA). Table 1 in the Statistical Appendix

1 A total of 623,871 surgical cases were performed during these 190,071 OR-days.

46

2

shows all direct and indirect (other word for ‘mediation’) relationships between the performance indicators that were assessed, as well as the specific sets of statistical analyses applied.

Simple linear regression analysis was used to identify direct relationships and corresponding R-Squared (R2) values between performance indicators. Multiple linear regression analysis was applied to assess the direct relationship (with corresponding R2 values) between the response variable raw utilization and the three predictor variables late start, turnover time and underutilized time. To justify the use of (multiple) linear regression analysis additional tests were performed to test the general assumptions25. R-Squared (R2) values can be interpreted as representing the percentage of variation in the dependent variable explained by variation in the independent variables. The higher the R2, the better the regression model fits the data.

To evaluate the indirect effect of these three indicators of ‘non-operative’ time on OR utilization, a mediation analysis was completed26, which investigates whether the effect of a predictor variable X on a response variable Y was influenced by a third predictor variable, known as a mediator variable M. The mediational effect in which X leads to Y through M is also called the indirect effect. The mediation analysis was conducted by the Baron and Kenny method26. See also Figure 1a. The indirect effects were translated into the following three hypotheses:

a. late start initiated an increase in turnover time;b. late start initiated a decrease in underutilized time;c. turnover time initiated a decrease in underutilized time.

3 RESULTS

The eight centers together realized a mean (± SD) raw utilization of 82% (± 16%) and a median of 87%, for all in-patient OR days: a total of 190,071 OR days on which 623,871 cases were performed from 2005 up to and including 2011. Mean (± SD) of late start 26 (± 40) minutes and median of 10 minutes; mean (± SD) turnover time of 33 (± 27) minutes and median of 25 minutes; mean (± SD) underutilized time of 65 (± 55) minutes with a median of 49 minutes.

Additional tests to check the general assumptions of regression analysis showed that linearity, as well as independence, were not violated. However, the assumptions regarding homoscedasticity and normal error distribution were violated. To correct for heteroscedasticity, several analyses were computed, of which all details are described in the statistical appendix. Based on these extra analyses, we conclude that the results in this study were not influenced by heteroscedasticity27. Therefore, we conclude that the analysis reported in this study was not influenced by heteroscedasticity27. Concerning normality of the error distribution the assumption, linear regression is considered robust against this assumption, particularly with large sample sizes (N ≥ 1,000)28-30, which was the case in this study.

47

Enh

ance

men

t Opp

ortu

nitie

s in

Ope

ratin

g R

oom

Util

izat

ion3.1 First set of analyses: direct effect of ‘non-operative’ time indicators on raw

utilizationThe empirical data of all UMCs showed that the later the start, the less raw utilization; correlation coefficient R= -0.524 (P< 0.01). Twenty-seven percent (R2=0.274)of raw utilization was explained by late start (P< 0.01). The longer the turnover time, the less raw utilization; correlation coefficient R= -0.378 (P< 0.01). Merely 14% (R2= 0.143) of raw utilization could be explained by turnover time (P< 0.01).

The more underutilized time, the less raw utilization; R= -0.639 (P< 0.01). Forty-one percent (R2= 0.409)of raw utilization can be attributed to underutilized time (P< 0.01). Based on multiple linear regression (P< 0.01) underutilized time showed the highest absolute β-value of -0.699 and thus the greatest negative influence, followed by late start (β= -0.500) and finally turnover time (β= -0.383). Overall, 87% (R2=0.867) of raw utilization was explained by the three indicators of ‘non-operative’ time together (P< 0.01).

3.2 Second set of analyses: interaction between late start, turnover time and underutilized timeData of all UMCs showed that late start and turnover time, as well as late start and underutilized time, had a significant positive relationship (P< 0.01). In other words: the later the start, the longer the turnover time, and the more underutilized time. Turnover time and underutilized time also showed a significant yet negative relationship, meaning that the longer the turnover time, the less underutilized time. These relationships were not strong, but statistically significant: based on low values of determination coefficients (R2) merely 1% of turnover time was explained by late start; 2% of underutilized time was explained by late start and 1% by turnover time.

3.3 Third set of analyses: indirect effect of ‘non-operative’ time indicators on OR utilizationMediation analysis investigated the mediational effects (X leads to Y through M), which are also named ‘indirect effects’. Table 1 shows the indirect effects that were evaluated (see also Table A in the Statistical Appendix) and the corresponding results on an overall level of all eight UMCs (complete dataset).

3.4 Hypothesis a: late start initiated an increase in turnover timeThis first hypothesis was confirmed based on the complete dataset. Data showed a partial indirect effect of turnover time on the relationship between late start and raw utilization. In other words: the later the start, the longer the turnover time, the less raw utilization. This indirect effect was small, but statistically significant. See also Figure 1b.

48

2

3.5 Hypothesis b: late start initiated a decrease in underutilized timeThis second hypothesis was rejected based on the complete dataset. Mediation analysis results demonstrated that late start had a direct and indirect (through underutilized time) negative influence on raw utilization. I.e. the later the start, the more underutilized time at the end of the day, the less raw utilization. This indirect effect was small, yet statistically significant.

3.6 Hypothesis c: turnover time initiated a decrease in underutilized timeBased on the complete dataset of all eight centers this final hypothesis was confirmed. Results showed that the longer the turnover time, the less the raw utilization. Mediation analysis revealed that the longer the turnover time, the less underutilized time at the end of the day, resulting in a little more raw utilization. This indirect effect was small, but statistically significant.

All statistical analyses were performed for all UMCs in total (the complete dataset) as well as per hospital separately. The results showed no significant differences between hospitals (no inter-hospital variability) and therefore the results from the collaborative were consistent across each university hospital.

49

Enh

ance

men

t Opp

ortu

nitie

s in

Ope

ratin

g R

oom

Util

izat

ion

Tabl

e 1.

Res

ults

of m

edia

tion

anal

ysis

A

naly

zed

med

iatio

n/in

dire

ct

effe

cts

Test

Valu

eP-

valu

eR

esul

ts (a

ll 8

cent

ers)

apr

edict

or v

aria

ble

late

sta

rt (X

)re

spon

se v

aria

ble

raw

utiliz

atio

n (Y

)m

edia

tor v

aria

ble

turn

over

tim

e (M

)

Sm

all,

yet s

igni

fican

t pa

rtial

indi

rect

effe

ct: t

he

late

r the

sta

rt, th

e lo

nger

th

e tu

rnov

er ti

me,

the

less

ra

w ut

ilizat

ion.

M =

i 1 + a

X (S

Ea)

M =

34

+ 0.

071X

(SE

a = 0

.003

)(P<

0.00

1)

Y

= i 2 +

cX

Y =

86 -

0.22

5X(P<

0.00

1)

Y

= i 3 +

c’X

+ b

M (S

Eb)

Y =

92 -

0.21

7X -

0.17

1M (S

Eb =

0.0

01)

(P<

0.00

1)

Sobe

l tes

tz-

valu

e =

a*b/

SQ

RT(

b2 *S

Ea2 +

a2 *

SE

b2 )S

obel

test

= -2

3.44

(P<

0.01

)

indi

rect

effe

cta*

b =

-0.0

12pa

rtial

med

iatio

n ef

fect

c’ =

-0.2

17

b

pred

ictor

var

iabl

e la

te s

tart

(X)

resp

onse

var

iabl

e ra

w ut

ilizat

ion

(Y)

med

iato

r var

iabl

e un

deru

tilize

d tim

e (M

)

Sm

all,

yet s

igni

fican

t pa

rtial

indi

rect

effe

ct: t

he

late

r the

sta

rt, th

e m

ore

unde

rutili

zed

time

at th

e en

d of

the

day,

the

less

ra

w ut

ilizat

ion.

M =

i 1 + a

X (S

Ea)

M =

62

+ 0.

171X

(SE

a = 0

.006

)(P<

0.00

1)

Y

= i 2 +

cX

Y =

86 -

0.22

5X(P<

0.00

1)

Y

= i 3 +

c’X

+ b

M (S

Eb)

Y =

92 -

0.21

4X -

0.17

0M (S

Eb =

0.0

01)

(P<

0.00

1)

Sobe

l tes

tz-

valu

e =

a*b/

SQ

RT(

b2 *S

Ea2 +

a2 *

SE

b2 )S

obel

test

= -2

8.11

(P<

0.01

)

indi

rect

effe

cta*

b =

-0.0

29pa

rtial

med

iatio

n ef

fect

c’ =

-0.2

14

c

pred

ictor

var

iabl

e tu

rnov

er ti

me

(X)

resp

onse

var

iabl

e ra

w ut

ilizat

ion

(Y)

med

iato

r var

iabl

e un

deru

tilize

d tim

e (M

)

Sm

all,

yet s

igni

fican

t pa

rtial

indi

rect

effe

ct: t

he

long

er th

e tu

rnov

er ti

me,

th

e le

ss u

nder

utiliz

ed

time

at th

e en

d of

the

day,

resu

lting

in a

little

mor

e ra

w ut

ilizat

ion.

M =

i 1 + a

X (S

Ea)

M =

62

- 0.1

38X

(SE

a = 0

.007

)(P<

0.00

1)

Y

= i 2 +

cX

Y =

90 -

0.19

6X(P<

0.00

1)

Y

= i 3 +

c’X

+ b

M (S

Eb)

Y =

96 -

0.21

0X -

0.19

2M (S

Eb =

0.0

01)

(P<

0.00

1)

Sobe

l tes

tz-

valu

e =

a*b/

SQ

RT(

b2 *S

Ea2 +

a2 *

SE

b2 )S

obel

test

= 1

9.61

(P<

0.01

)

indi

rect

effe

cta*

b =

0.02

6pa

rtial

med

iatio

n ef

fect

c’ =

-0.2

10

50

2

Fig

ure

1a

. M

ed

iatio

n a

na

lysis

(th

eo

ry)

Ind

ep

en

den

t / P

red

icto

r V

ari

ab

le (

X)

Dep

en

den

t / O

utc

om

e

Vari

ab

le (

Y)

Med

iato

rV

ari

ab

le

path

ain

direct eff

ectbetw

een

independent

and m

edia

tor

variable

path

bin

direct eff

ectbetw

een

media

tor

and d

ependent

variable

path

cdirect eff

ect betw

een

independent

and d

ependent

variable

path

c'

media

tor

variable

reduces

(part

ial m

edia

tion e

ffect)

the

link b

etw

een t

he independent

and d

ependent

variable

If c

‘ <

c b

ut

c'>

0: d

ata

su

gg

estp

art

ial m

ed

iati

on

. If

c‘ =

0: d

ata

su

gg

est to

tal m

ed

iati

on

(i.e

., th

e m

ed

iato

r co

mp

lete

ly

acco

un

ts fo

r th

e r

ela

tio

nsh

ip b

etw

ee

n th

e in

de

pe

nd

en

t a

nd

de

pe

nd

en

t va

ria

ble

).

Ca

lcu

late

a z

sc

ore

in

ord

er

to e

sta

blis

h th

e s

ign

ific

ance

of

the

me

dia

tio

n

(diffe

ren

ce

be

twe

en

c a

nd

c').

:

=

∗

∗

+(

∗

)

z =

a*b

/SQ

RT

(b2*s

ea

2+

a2*s

eb

2)

If z

sco

re >

1.9

6 th

e m

ed

iatio

n is

sig

nific

ant a

t p =

.0

5If

z s

co

re >

2.4

6 th

e m

ed

iatio

n is

sig

nific

ant a

t p =

.0

1Fi

gure

1a.

Med

iatio

n an

alys

is (t

heor

y)

51

Enh

ance

men

t Opp

ortu

nitie

s in

Ope

ratin

g R

oom

Util

izat

ion

late

sta

rt (

X)

raw

uti

liza

tio

n(Y

)

turn

ove

rti

me

(M

)

path

ain

direct

eff

ect

a=

0.0

71*

(SE

a=

0.0

03)

path

bin

direct

eff

ect

b=

-0.1

71*

(SE

b=

0.0

01)

path

cdirect

eff

ect

c=

-0.2

25*

part

ial m

edia

tion e

ffect

c'=

-0.2

17

Fig

ure

1b.

Media

tion a

naly

sis

(one p

art

of

the r

esults)

Note

* =

P<

0.0

01

Sobel z-v

alu

ein

direct

eff

ect a*b

-23.4

4(P

< 0

.001)

-0.0

12

Figu

re 1

b. M

edia

tion

anal

ysis

(one

par

t of t

he re

sults

)N

ote

* = P

<0.0

01

52

2

4. DISCUSSION

The findings from this Dutch, nationwide, multicenter study show that in a university hospital environment, underutilized OR time at the end of the day has the strongest influence on raw utilization, followed by late start and turnover time. A direct negative relationship between all three indicators (late start, turnover time, underutilized time) and raw utilization was found.

Late start and underutilized time showed a positive relationship whereas a negative relationship between turnover time and underutilized time was observed. Based on our findings, late start, turnover time and underutilized time were ‘stand-alone’ aspects with an important direct influence on raw utilization and only a minor influence on each other. We were unable to verify the reported ‘trickle down’ effect10, caused by late start and resulting in an increased delay as the day progresses. The interaction between the three ‘non-operative’ indicators, as well as their indirect effects on raw utilization were inconsequential. The findings from this study are important for hospital management and surgical teams, since they clearly suggest that improving OR utilization should be focused on reducing the amount of underutilized time at the end of the day.

Potential solutions and interventions to address the issue of underutilized OR time are: improving the prediction of the total procedure time of surgical cases; altering the sequencing of scheduled operations and altering patient cancellation policies. These interventions are discussed below.

When an operation takes longer than predicted, subsequent operations may need to be postponed or even cancelled. When the actual time of an operation is shorter than predicted, the OR remains unused at the end of the day. Both situations are unwelcome and could lead to suboptimal utilization of the OR2. The reduction of underutilized time may be possible by improving the prediction of the total procedure time of operations and thus improving OR scheduling. Scheduling surgical procedures is complex because predicting total procedure time entails several elements subject to variability, including the two main components: surgeon-controlled time and anesthesia-controlled time, each with a considerable random chance component23. The efficiency of OR scheduling is greatly improved with better ability to accurately predict the time needed for all components of care for each surgical case2, 23, 31-36.

An alternative method to enhance OR scheduling and generate reductions in underutilized time as well as overtime, is to alter the sequencing of scheduled operations. Prior studies suggested that it is better to schedule short procedures before long operations and alternating between the two, which can limit the variability in case duration and can make predictability more accurate37-39.

A reduction of underutilized time might also be possible by altering patient cancellation policies2, 40, 41, 42. A practice applied in many Dutch (university) hospitals is a ‘zero tolerance for overtime’ policy, because OR management presumes it is more economically profitable to finish

53

Enh

ance

men

t Opp

ortu

nitie

s in

Ope

ratin

g R

oom

Util

izat

ionthe daily OR caseloads during ‘regular’ hours than to create overtime41, 42. A consequence of

this policy may be that a patient scheduled in the final allocated hours (or late afternoon) will be cancelled last-minute to avoid overtime. This leads to immaterial damage concerning postponed or cancelled patients and to financial losses for the hospital concerning under-utilization of scarce OR capacity. Because all OR personnel in Dutch UMCs are contracted and paid for at least 8 hours on each day worked, underutilized time leads to economic losses for the hospital due to these fixed labor costs. Tessler42 and Stepaniak41, however, showed in their previous work that it is more cost-effective to proceed with an operation after regular hours than to cancel this operation. Overtime does have a financial effect owing to the payment of overtime wages beyond the regular rate for 36 hours a week (in Dutch UMCs). Working overtime can also have a negative influence on job satisfaction of registered nurses and is considered a reason to change their employment status43, 44. To better absorb the consequences of underutilized time as well as overtime, one option could be to employ OR personnel on a flexible basis adjusted to patient needs, as suggested in previous research6, 12, 45, 46.

The direct and indirect effects of ‘non-operative’ time on raw utilization is worthwhile studying because former research concluded that late start can cause a ‘trickle down’ effect resulting in an increased delay (of e.g. turnover time) as the day progresses, potentially affecting the rest of the scheduled patients10. Our research, however, implicates that the indirect effects of late start through turnover time and underutilized time do not have a major impact on raw utilization. Therefore, these recent results reconfirm several earlier studies that resources spent solely on trying to achieve on time starts of scheduled first cases will not considerably improve OR utilization or productivity8, 11, 47-49, and the ‘trickle down’ effect has not yet been verified6, 8, 21, 49. Our study reveals that a late start can be caught up throughout the rest of the OR-day, either during operative time or due to a quicker turnover. Future research should investigate this specific subject to reveal its principles.

The findings in this study are subject to at least two limitations. First, data for this study were gathered in tertiary referral centers only, and therefore generalization of the findings to general hospitals may be limited. Our earlier research showed that the complexity of surgical cases as well as their duration is generally greater than in general hospitals23. This level of complexity of the patient case mix in UMCs can make it more difficult to accurately predict their duration and hamper efficient scheduling. Uncertainty, variability and length in the duration of surgery contribute to the difficulty of scheduling1, 50, which may lead to either much underutilized time or unwanted overtime at the end of the day. One can imagine that in general hospitals with less complex patients, shorter case durations and the attendant reduced variability, case durations can be more accurately predicted. This, in turn will result in more effective scheduling with efficient use of OR resources2 (with less underutilized time and less overtime). One can also imagine that in general hospitals with smaller OR facilities (e.g. with a total of up to 10 ORs) turnover times can be shorter than in UMCs with large OR

54

2

facilities (20 or more ORs). Smaller facilities deal with a shorter patient transport time from ward to the holding area and from the holding area to the operating room.

All statistical analyses were performed for all UMCs in total (complete dataset) as well as per hospital separately. The results showed no significant inter-hospital variability and the same conclusions were justified for all UMCs. In other words, the results from the collaborative were consistent across each hospital. This may argue for increased generalizability of our results in a university hospital OR environment.

Second, the study does not consider all performance indicators relevant to ‘the end of an OR-day’, because the study focusses solely on ‘non-operative’ time and raw utilization. The end of one OR-day balances between either underutilized OR time or overtime (also called over-utilized OR time), along with the potential cancellations of elective surgical cases. Reasons for cancellation should also be involved in detail. Avoidable cancellations as well as unavoidable cancellations are relevant in this respect. Investigation of the indirect effects of late start and turnover time should evaluate the relationship between start and finish times, considering all relevant indicators. Nonetheless, earlier research has showed that the most common factor for cancellation is lack of availability of OR time51.

Future research requires a clear and unambiguous definition of a cancellation and its reasons. Moreover, future research should involve the specific registration of which operations were overruled, elective or emergency cases. This information is generally registered on paper or Excel sheet, separate from digital clinical records or Hospital Information System, and therefore problematic to pool with databases similar to the one used in this study.

Even though this research concentrates on efficient utilization of scarce and expensive OR capacity, improving utilization regularly concurs with improving quality and patient safety. One Dutch UMC implemented preoperative cross-functional teams (CFT’s) based on a socio-technical design, responsible for OR scheduling, with the aim to increase efficient utilization of OR capacity and enhance patient safety52. The CFT meets once a week to discuss the OR schedule of the following week and to evaluate the OR performance of the previous week, in terms of utilization, cancellations and all relevant issues concerning optimal planning, performance and safety. This approach led to an organizational learning effect and demonstrated that OR utilization as well as patient safety can be improved by allowing the individual healthcare workers to function as a team. Although their study is preliminary, it now serves as a starting point for more comprehensive studies in cooperation with the Dutch OR Benchmarking Collaborative to expand these initial findings52.

In summary, we suggest that hospital management and surgical teams direct scarce financial means and efforts on decreasing underutilized time, because it has the strongest influence on OR utilization. This advice is supported by an extensive, nationwide and diverse OR dataset of eight UMCs and the relationships found. Reduction of underutilized time can be accomplished by engaging the challenge to enhance OR scheduling.

55

Enh

ance

men

t Opp

ortu

nitie

s in

Ope

ratin

g R

oom

Util

izat

ionREFERENCES



3. Peltokorpi A. How do strategic decisions and operative practices affect operating room productivity? Health Care Manag Sci;14:370-82.

4. Wong J, Khu KJ, Kaderali Z, Bernstein M. Delays in the operating room: signs of an imperfect system. Can J Surg 2010;53:189-95.

5. Van Veen-Berkx E, Elkhuizen SG, Kalkman CJ, Buhre WF, Kazemier G. Successful interventions to reduce first-case tardiness in Dutch university medical centers: Results of a nationwide operating room benchmark study. Am J Surg 2014.

6. Dexter F, Epstein RH. Typical savings from each minute reduction in tardy first case of the day starts. Anesth Analg 2009;108:1262-7.

7. Fezza M, Palermo GB. Simple solutions for reducing first-procedure delays. AORN J 2011;93:450-4.

8. Pandit JJ, Abbott T, Pandit M, Kapila A, Abraham R. Is ‘starting on time’ useful (or useless) as a surrogate measure for ‘surgical theatre efficiency’? Anaesthesia 2012;67:823-32.

9. Wachtel RE, Dexter F. Reducing tardiness from scheduled start times by making adjustments to the operating room schedule. Anesth Analg 2009;108:1902-9.

10. Wright JG, Roche A, Khoury AE. Improving on-time surgical starts in an operating room. Can J Surg 2010;53:167-70.

11. Mazzei WJ. Operating room start times and turnover times in a university hospital. J Clin Anesth 1994;6:405-8.

12. Overdyk FJ, Harvey SC, Fishman RL, Shippey F. Successful strategies for improving operating room efficiency at academic institutions. Anesth Analg 1998;86:896-906.

13. Harders M, Malangoni MA, Weight S, Sidhu T. Improving operating room efficiency through process redesign. Surgery 2006;140:509-14; discussion 14-6.

14. Strum DP, Vargas LG, May JH. Surgical subspecialty block utilization and capacity planning: a minimal cost analysis model. Anesthesiology 1999;90:1176-85.

15. Agnoletti V, Buccioli M, Padovani E, et al. Operating room data management: improving efficiency and safety in a surgical block. BMC Surg 2013;13:7.

16. Van Houdenhoven M, Hans EW, Klein J, Wullink G, Kazemier G. A norm utilisation for scarce hospital resources: evidence from operating rooms in a Dutch university hospital. J Med Syst 2007;31:231-6.

17. Donham RT. Defining measurable OR-PR scheduling, efficiency, and utilization data elements: the Association of Anesthesia Clinical Directors procedural times glossary. International anesthesiology clinics 1998;36:15-29.

18. Glossary of times used for scheduling and monitoring of diagnostic and therapeutic procedures. AORN J 1997;66:601-6.

19. van Veen-Berkx E, Elkhuizen SG, Kalkman CJ, Buhre WF, Kazemier G. Successful interventions to reduce first-case tardiness in Dutch university medical centers: Results of a nationwide operating room benchmark study. Am J Surg 2014;207:949-59.

56

2

20. Weinbroum AA, Ekstein P, Ezri T. Efficiency of the operating room suite. Am J Surg 2003;185:244-50.

21. Dexter F, Macario A. When to release allocated operating room time to increase operating room efficiency. Anesth Analg 2004;98:758-62, table of contents.

22. Kazemier G, Van Veen-Berkx E. Comment on “Identification and use of operating room efficiency indicators: the problem of definition”. Can J Surg 2013;56:E103-4.

23. Van Veen-Berkx E, Bitter J, Elkhuizen SG, et al. The Influence of Anesthesia-Controlled Time on Operating Room Scheduling in Dutch University Medical Centers. Can J Anesth 2014;In press.

24. Van Houdenhoven M, van Oostrum JM, Hans EW, Wullink G, Kazemier G. Improving operating room efficiency by applying bin-packing and portfolio techniques to surgical case scheduling. Anesth Analg 2007;105:707-14.

25. Bruce N, Pope D, Stanistreet D. Quantitative Methods for Health Research:A Practical Interactive Guide to Epidemiology and Statistics. 2nd ed. West Sussex, England: John Wiley and Sons Ltd.; 2008.

26. Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. Journal of personality and social psychology 1986;51:1173-82.

27. Hayes AF, Cai L. Using heteroskedasticity-consistent standard error estimators in OLS regression: an introduction and software implementation. Behav Res Methods 2007;39:709-22.

28. Agresti A, Finlay B. Statistical Methods for the Social Sciences. 4 ed: Pearson Prentice Hall; 2009.

29. De Heus P, Van der Leeden R, Gazendam B. Toegepaste Data-analyse. 7 ed: Reed Business ‘s-Gravenhage, the Netherlands; 2008.

30. Lumley T, Diehr P, Emerson S, Chen L. The importance of the normality assumption in large public health data sets. Annual review of public health 2002;23:151-69.

31. Dexter F, Dexter EU, Ledolter J. Influence of procedure classification on process variability and parameter uncertainty of surgical case durations. Anesth Analg 2010;110:1155-63.

32. Strum DP, Sampson AR, May JH, Vargas LG. Surgeon and type of anesthesia predict variability in surgical procedure times. Anesthesiology 2000;92:1454-66.

33. Ehrenwerth J, Escobar A, Davis EA, et al. Can the attending anesthesiologist accurately predict the duration of anesthesia induction? Anesth Analg 2006;103:938-40.

34. Wright IH, Kooperberg C, Bonar BA, Bashein G. Statistical modeling to predict elective surgery time. Comparison with a computer scheduling system and surgeon-provided estimates. Anesthesiology 1996;85:1235-45.

35. Dexter F, Macario A. Applications of information systems to operating room scheduling. Anesthesiology 1996;85:1232-4.

36. Pandit JJ, Tavare A. Using mean duration and variation of procedure times to plan a list of surgical operations to fit into the scheduled list time. Eur J Anaesthesiol 2011;28:493-501.

37. Denton B, Viapiano J, Vogl A. Optimization of surgery sequencing and scheduling decisions under uncertainty. Health Care Manag Sci 2007;10:13-24.

38. Iser JH, Denton BT, King RE. Heuristics for Balancing Operating Room and Post-Anesthesia Resources Under Uncertainty. Proceedings of the 2008 Winter Simulation Conference 2008.

39. Lebowitz P. Schedule the short procedure first to improve OR efficiency. AORN J 2003;78:651-4, 7-9.

40. Tyler DC, Pasquariello CA, Chen CH. Determining optimum operating room utilization. Anesth Analg 2003;96:1114-21.

57

Enh

ance

men

t Opp

ortu

nitie

s in

Ope

ratin

g R

oom

Util

izat

ion

41. Stepaniak PS, Mannaerts GH, de Quelerij M, de Vries G. The effect of the Operating Room Coordinator’s risk appreciation on operating room efficiency. Anesth Analg 2009;108:1249-56.

42. Tessler MJ, Kleiman SJ, Huberman MM. A “zero tolerance for overtime” increases surgical per case costs. Can J Anaesth 1997;44:1036-41.

43. Shader K, Broome ME, Broome CD, West ME, Nash M. Factors influencing satisfaction and anticipated turnover for nurses in an academic medical center. J Nurs Adm 2001;31:210-6.

44. Strachota E, Normandin P, O’Brien N, Clary M, Krukow B. Reasons registered nurses leave or change employment status. J Nurs Adm 2003;33:111-7.

4 5. Dexter F, Macario A, Manberg PJ, Lubarsky DA. Computer simulation to determine how rapid anesthetic recovery protocols to decrease the time for emergence or increase the phase I postanesthesia care unit bypass rate affect staffing of an ambulatory surgery center. Anesth Analg 1999;88:1053-63.

4 6. Macario A, Dexter F. Effect of compensation and patient scheduling on OR labor costs. AORN J 2000;71:860, 3-9.

4 7. Dexter EU, Dexter F, Masursky D, Garver MP, Nussmeier NA. Both bias and lack of knowledge influence organizational focus on first case of the day starts. Anesth Analg 2009;108:1257-61.

4 8. Dexter F. Is time on first-case starts well spent? OR Manager 2010;26:22-4.

4 9. McIntosh C, Dexter F, Epstein RH. The impact of service-specific staffing, case scheduling, turnovers, and first-case starts on anesthesia group and operating room productivity: a tutorial using data from an Australian hospital. Anesth Analg 2006;103:1499-516.

5 0. Cardoen B, Demeulemeester E, Belien J. Operating room planning and scheduling: A literature review. European Journal of Operational Research 2010;201:921-32.

5 1. Kumar R, Gandhi R. Reasons for cancellation of operation on the day of intended surgery in a multidisciplinary 500 bedded hospital. Journal of anaesthesiology, clinical pharmacology 2012;28:66-9.

5 2. Bitter J, van Veen-Berkx E, Gooszen HG, van Amelsvoort P. Multidisciplinary teamwork is an important issue to healthcare professionals. Team Performance Management 2013;19:263-78.

58

2

STATISTICAL APPENDIXData analysis was performed using SPSS Statistics 20 (IBM SPSS Statistics for Windows, version 20.0, IBM Corp. Released 2011.; Armonk, NY, USA).

Assumptions (multiple) linear regression analysisAdditional tests were performed to test the four general assumptions with the aim to justify the use of (multiple) linear regression analysis1:

1. Linearity of the relationship between dependent and independent variables;2. Assumption of homoscedasticity (the errors/residuals have the same variance);3. Assumption of independence (the errors are independent of each other);4. Assumption of normality (the errors are normally distributed).

To prevent ‘over fitting’ of the regression analysis, collinearity statistics consisting of Tolerance and VIF-values for all three independent variables, were computed2. Furthermore, additional tests regarding influence diagnostics on regression coefficients, such as Cook’s Distance and DFFITS, were performed. Extreme values i.e. outliers were checked to prevent them from distorting the estimates of regression coefficients2.

RESULTS

Additional tests to check the general assumptions of regression analysis showed that linearity, as well as independence, were not violated. However, the assumptions regarding homoscedasticity and normal error distribution were violated. To correct for heteroscedasticity, bias-corrected and accelerated (BCa) confidence intervals were computed. These bootstrapped confidence intervals for the regression coefficients were considered narrow. Moreover, Generalized Linear Models were applied and heteroscedasticity-robust standard errors (also called Huber-White standard errors)3 were calculated. Comparing these heteroscedasticity-robust standard errors with the original regression output showed the same statistically significant results for all regression estimates, as well as minor standard errors. Therefore, we conclude that the analysis reported in this study was not influenced by heteroscedasticity4. Concerning normality of the error distribution the assumption, linear regression is considered robust against this assumption, particularly with large sample sizes (N ≥ 1,000)5-7, which was the case in this study.

Collinearity statistics regarding all three independent variables showed Tolerance values as well as VIF-values close to 1 and therefore indicate that multicollinearity is not a problem in this study2. Cook’s Distance scores had a minimum value of 0.000 and a maximum value of 0.029, and were not larger than the threshold of “1”2. Standardized DFFITS showed a minimum value of -0.017 and a maximum value of 0.015, and were not larger than the threshold of “2”2.

59

Enh

ance

men

t Opp

ortu

nitie

s in

Ope

ratin

g R

oom

Util

izat

ionThe independent data management center removed extreme values from the dataset. No

data points used in this study were considered influential.

Mediation analysis by Baron and KennyTo evaluate the indirect effect of these three indicators of non-operative time on OR utilization, a mediation analysis was completed8, which investigates whether the effect of a predictor variable X on a response variable Y is influenced by a third predictor variable, known as a mediator variable M. The mediation analysis was conducted by the Baron and Kenny method8. This method contained four different steps, see Table 1, 4th set of analyses, a-d.

First, the direct relationship between variables X and Y, as well as the direct relationship between variables X and M were tested with simple linear regression analysis. Second, a multiple regression analysis of variables M and X on variable Y was applied to check whether the direct relationships last when X and M both influence Y. Third, a multiple regression analysis was applied to assess the influence of M and X on Y. At last, significance was determined for all preceding steps to confirm mediation. Partial mediation means that variable X has a direct effect on variable Y, and that variable X has an indirect effect on variable Y through variable M (Figure 1a in Manuscript). Full mediation means that variable X does not have a direct effect on variable Y, but variable X purely has an indirect effect on variable Y through variable M (Figure 1a in Manuscript).

To explain the “a, b, c, c’ paths” Figure 1 contains two parts a) and b). Figure 1a) explicates the theory of mediation analysis and the Sobel test equation by the Baron and Kenny method8. Figure 1b) explicates the actual coefficients of the recent study above the arrows linking the variables. The example concentrates on one of the total four partial mediation effects analyzed, see Table 1 “Overview of statistical analyses applied”, 4th set of analyses, first bullet: predictor variable late start (X) response variable raw utilization (Y) mediator variable turnover time (M).

60

2

RESULTS

For all results of the mediation analysis by Baron and Kenny, see Manuscript.

Table A. Overview of statistical analyses applied

Sets of analyses

Relationship Statistical analysis

1. Direct linear relationship regarding raw utilization:• Predictor (independent) variable late start

response (dependent) variable raw utilization• Predictor variable turnover time response variable

raw utilization• Predictor variable underutilized time response

variable raw utilization

Simple linear regression analysis

2. Direct linear relationship regarding raw utilization:• Predictor variables late start, turnover time and

underutilized time response variable raw utilization

Multiple linear regression analysis including interaction effects

3. Direct linear relationship regarding non-operative time:• Predictor variable late start response variable

turnover time• Predictor variable late start response variable

underutilized time• Predictor variable turnover time response variable

underutilized time

Simple linear regression analysis

4. Partial mediation effect:• Predictor variable late start (X) response variable

raw utilization (Y) mediator variable turnover time (M)

• Predictor variable late start (X) response variable raw utilization (Y) mediator variable underutilized time (M)

• Predictor variable turnover time (X) response variable raw utilization (Y) mediator variable underutilized time (M)

Mediation effect analysis was calculated by the Baron and Kenny method:a. Simple linear regression with X predicting M to

test for path aM = i1 + aX

b. Simple linear regression with X predicting Y to test for path c Y = i2 + cX

c. Multiple regression analysis with both X and M predicting Y to test for path b (M Y) and for path c’ (X & M Y)Y = i3 + c’X + bM

d. Significance test applied to a, b and c: P-value checked on a and b. Sobel test checked on c mediation is confirmed when all parts of the analysis are significant

i1, i2 and i3 represent the intercepts for each model

61

Enh

ance

men

t Opp

ortu

nitie

s in

Ope

ratin

g R

oom

Util

izat

ion

REFERENCES STATISTICAL APPENDIX

1. Bruce N, Pope D, Stanistreet D. Quantitative Methods for Health Research:A Practical Interactive Guide to Epidemiology and Statistics. 2nd ed. West Sussex, England: John Wiley and Sons Ltd.; 2008.


3. Huber PJ. The behavior of maximum likelihood estimat es under nonstandard conditions. Berkeley, California: University of California Press; 1967.

4. Hayes AF, Cai L. Using heteroskedasticity-consistent standard error estimators in OLS regression: an introduction and software implementation. Behav Res Methods 2007;39:709-22.



7. Lumley T, Diehr P, Emerson S, Chen L. The importance of the normality assumption in large public health data sets. Annual review of public health 2002;23:151-69.

8. Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. Journal of personality and social psychology 1986;51:1173-82.

3The Influence of Anesthesia-Controlled Time

on Operating Room Scheduling in


Elizabeth van Veen-Berkx, MScJustin Bitter, MScSylvia G. Elkhuizen, PhD Wolfgang F. Buhre, MD, PhDCor J. Kalkman, MD, PhDHein G. Gooszen, MD, PhDGeert Kazemier, MD, PhD for the Dutch Operating Room Benchmarking Collaborative

CanadianJournalofAnesthesia.2014.61(6):524-32.

64

3

ABSTRACT

Background: Predicting total procedure time (TPT) entails several components subject to variability, including the two main constituents: surgeon-controlled time (SCT) and anesthesia-controlled time (ACT). This study explores the effect of ACT on TPT as a proportion of TPT as opposed to a fixed number of minutes. The goal is to enhance the prediction of TPT and improve OR scheduling.

Methods: Data from six University Medical Centers (UMCs) over seven consecutive years (2005-2011) were included: a total of 330,258 in-patient elective operations. Based on the actual ACT and SCT the revised prediction of TPT was determined as SCT x 1.33. Differences between actual and predicted total procedure times were calculated for the two methods of prediction.

Results: An improvement in the predictability of TPT showed when the scheduling of procedures was based on predicting ACT as a proportion of SCT.

Conclusions: Efficient OR management demands the accurate prediction of the times needed for all components of care, including SCT and ACT, for each surgical procedure. Supported by an extensive dataset from six UMCs, we advise grossing up the SCT by 33% to account for ACT (revised prediction of TPT = SCT x 1.33), as opposed to employing a fixed number of minutes methodology for ACT. This recommendation will improve OR scheduling, which might result in the reduction of over-utilized OR time and case cancellations, and therefore in more efficient use of limited OR resources.

65

Ane

sthe

sia-

cont

rolle

d Ti

me

and

Ope

ratin

g R

oom

Sch

edul

ing

ABBREVIATIONS

OR Operating RoomUMC University Medical CenterUMCs University Medical Centers TPT Total Procedure TimeSCT Surgeon-Controlled TimeACT Anesthesia-Controlled TimepTPT Predicted TPT in the current prediction method (= pSCT + pACT)pSCT Surgeon’s prediction SCT in the current prediction methodpACT Fixed ACT in the current prediction methodaTPT Actual Total Procedure Time (= aSCT + aACT)aSCT Actual Surgeon-Controlled TimeaACT Actual Anesthesia-Controlled TimerpACT Revised predicted Anesthesia-Controlled TimerpTPT Revised predicted Total Procedure Time (= rpACT + aSCT)

66

3

INTRODUCTION

Operative surgical care is a costly multiprofessional activity1. Economic efficiency with regards to operative services is fundamentally tied to the efficient use of operative time, which in turn is tied to precision in scheduling surgical cases.

Among the hospitals participating in this study, approximately 60% of admitted patients received operative surgical care. Additionally, among the study hospitals 40% of revenues and expenses were related to operative surgical care emphasizing the importance of economic and operational efficiency. Optimal scheduling of operating rooms (ORs) is one way to achieve effective and efficient use of their capacity.

Scheduling surgical procedures is a complex process. One of the difficulties associated with the development of accurate OR schedules, is the uncertainty inherent to surgical procedures. Surgery duration variability represents deviations between the predicted and actual procedure time. Variability in the procedure time of operations complicates surgical scheduling and reduces operational efficiency2,3. When a procedure takes longer than predicted, this may lead to procedures being postponed or cancelled. When the time used is shorter than predicted, valuable operating room time is wasted. Both of these situations are undesirable and contribute to suboptimal use of the OR resources4.

Optimal scheduling depends on reliable predictions concerning the time needed for elective procedures are available. Total procedure time (TPT) is defined as the time from entry of the patient into the OR until leaving it. In this study the term “total procedure time” characterizes one “operation” and refers to one “session”. Predicting TPT is challenging because it entails several elements subject to variability, such as room setup and takedown, patient positioning, prepping and draping, as well as the two principal components: surgeon-controlled time (SCT) and anesthesia-controlled time (ACT).

If TPT could be better predicted, this could lead to a reduction in over-utilized OR time and fewer case cancellations2,5,11. While progress in OR scheduling methodology has been made over the past years, opportunities for improvement in this area of research still remain. Previous studies have examined methods to predict TPT and have aimed to develop predictive tools by statistical modeling of procedure times3,4,6-14. Most of these studies have focused on simulation of data, mathematical modeling, the selection of procedures from a few surgical departments or consideration of a single procedure. Although the number of modeling papers has grown substantially over recent years, few have reported on the outcomes of implementation of these models. Therefore, the value of implementing the suggested model remains unassessed. The validation of various models that have been identified remains to be published15.

Multicenter studies employing an extensive empirical dataset of OR scheduling systems are scarce. In this paper, we present empirical data from six different University Medical

67

Ane

sthe

sia-

cont

rolle

d Ti

me

and

Ope

ratin

g R

oom

Sch

edul

ing

Centers (UMCs) in the Netherlands over seven years. In these centers the time allocated on an OR schedule for a particular operation is the sum of ACT and SCT. The former has been considered to be a fixed number of minutes regardless of the surgical time. This study attempts to determine if this approach survives scrutiny or if an alternative approach to allocating ACT better predicts TPT to be allocated on a schedule. The goal is to enhance prediction of TPT and improve OR scheduling.

METHODS

Operating room departments of all eight UMCs in the Netherlands established a nationwide benchmarking collaboration in 2005, which is still active16,17. The objective of the collaboration is to improve OR performance by learning from each other through exchanging best practices. Each UMC provides data records for all surgical cases performed to a central OR Benchmark database. This extensive database, presently comprising more than one million records of surgical cases, is used to calculate key performance indicators related to utilization of OR capacity and to perform research on OR scheduling issues. An independent data management center enters the longitudinal data collection in the central OR Benchmark database. This center provides professional expertise facilitating the processing of data records and performing reliability checks before data are ready for analysis.

The central OR Benchmark database with a total of 940,381 cases consisted of records of all surgical cases performed at eight UMCs over a seven-year period from 2005 up to and including 2011. To define a consistent dataset for analysis, all non-elective (emergency) cases, surgical departments with a caseload in the OR of less than 100 per year, cases for which the registration of a specific surgical department was missing, and all outpatient cases, were excluded. In Dutch UMCs outpatient surgical cases are allocated to a specific organizational OR unit (a separate ‘day surgery center’). The outpatient surgery workflow varies from the in-patient surgery workflow and has a different planning methodology. If the OR department of a UMC is divided into a main location and sub locations such as a Children’s Hospital, Cancer Center or Thorax Center, the main (largest) in-patient OR location was included, because these sub locations also have a different planning methodology. During the session, OR nursing staff prospectively registered (electronically in the hospital information system in each UMC) the times for each case occupying the OR, and the surgeon and anesthesiologist in charge validated the times after completion of the session. Four time intervals were registered in the central OR benchmark database:

1. Total Procedure Time (predicted Total Procedure Time as well as actual Total Procedure Time),2. Anesthesia Induction Time (actual time),

68

3

3. Surgeon-Controlled Time (actual time),4. Anesthesia Emergence Time (actual time).

Anesthesia-controlled time is the sum of 2 and 4. Total procedure time consists of 2, 3 and 4; in other words ACT plus SCT. ACT was defined, according to Dexter18, as “the sum of the time starting when the patient enters the OR to the time when positioning or skin preparation can begin (2) plus the time starting when the surgical dressing is completed and ending when the patient leaves the OR” (4). SCT was also defined according to Dexter18, as “the time starting when patient positioning and/or skin preparation can begin to when surgical dressing is completed”. See Figure 1.

Figure 1. Total procedure time is subdivided into anesthesia induction time, surgeon-controlled time, and anesthesia emergence time. The sum of induction time and emergence time is anesthesia-controlled time. Predicting total procedure time entails several elements subject to variability, including the two main components: surgeon-controlled time and anesthesia-controlled time.

In the current prediction method, before each procedure, the surgeon’s prediction of SCT was routinely determined. A fixed time period of 20 minutes (general anesthesia) or 40 minutes (regional anesthetic technique) for ACT (pACT) was added to the surgeon’s prediction (pSCT). This provided the predicted TPT (pTPT), which was used for OR scheduling. The actual TPT (aTPT), actual ACT (aACT) and actual SCT (aSCT) were registered in the database. The difference between predicted and actual TPT was assessed.

Data from six UMCs and seven consecutive years (2005 – 2011) were included. Two UMCs were excluded because these centers were not able to record the predicted TPT due to the non-availability of an adequate recording system. Considering purely in-patient and elective operations, a total of 330,258 operations from the six UMCs were subjected to statistical analysis. First, the status quo of the relationship between the independent variable ‘predicted TPT’ and the dependent variable ‘actual TPT’ was assessed. Second, the proportion of TPT

Startsurgical case

Total Procedure Time (TPT) (one session)

Anesthesia Induction Time

Surgeon-Controlled Time (SCT)Anesthesia

Emergence Time

End case/start next case

Anesthesia-Controlled Time (ACT)


Surgeon-Controlled TimeAnesthesia

Emergence Time

69

Ane

sthe

sia-

cont

rolle

d Ti

me

and

Ope

ratin

g R

oom

Sch

edul

ing

attributed to ACT was calculated, as well as the ratio SCT/ACT. Third, based on this ratio, a ‘revised predicted ACT (rpACT)’ was computed and TPT was revised predicted based on this rpACT plus the surgeon’s prediction of SCT. Finally, the difference between revised predicted TPT (rpTPT) and actual TPT was assessed. Table 1 illustrates the conducted calculations.

Table 1. Conducted calculations

Total Procedure Time = ACT + SCT (ACT = induction time + emergence time)

330,258 in-patient elective operations were analyzed. The ratio SCT/ACT showed a mean (SD) of 3.61 (2.91) and a median of 2.90. These results indicate that aSCT is approximately three times greater than aACT. The mean of 3.61 was rounded down to 3, to correct for the influence of data outliers, because the variables used in this study were not normally distributed (TPT, SCT and ACT). Hence: (aACT/aTPT) x 100 = 25% and (aACT/aSCT) x 100 = 33%.

SCT = 3 x ACT if SCT = 3 x ACT then ACT = SCT/3 or ACT = SCT x 0.33

Revised prediction strategy: rpACT = pSCT x 0.33rpTPT = pSCT + (SCT x 0.33) orrpTPT = pSCT x 1.33

Suppose: pSCT =Then: rpACT = pSCT x 0.33And: rpTPT = pSCT + (pSCT x 0.33)Or: rpTPT = pSCT x 1.33

200 minutes200 x 0.33 = 66 minutes200 + (200 x 0.33) = 266 minutes200 x 1.33 = 266 minutes

ACT = anesthesia-controlled time; rpACT = revised predicted anesthesia-controlled time; SCT = surgeon-controlled time; pSCT = surgeon’s prediction SCT in the current prediction method; TPT total procedure time

Statistical analysisData analysis was performed using SPSS Statistics 19. Ordinary Least-Squares regression analysis was used to determine the regression line of aTPT on pTPT, as well as the regression line of aTPT on rpTPT. Scatter plots of actual versus predicted TPT and prediction error versus predicted TPT were constructed. Finally, stacked histograms of the prediction errors (current prediction method and revised prediction strategy) were assembled.

RESULTS

A total of 330,258 in-patient elective operations (2005-2011) were selected for inclusion in the study. The six centers demonstrated a mean (SD) aTPT of 158 (119) minutes and a median of 124 minutes, for all in-patient elective operations ((Table 2). The mean (SD) aSCT was 121 (106) minutes and the median aSCT was 90 minutes. The mean (SD) aACT was 37 (22) minutes and the median aACT was 31 minutes.

70

3

Table 2. Descriptive statistics of actual Total Procedure Time, actual Anesthesia-Controlled Time and actual Surgeon-Controlled Time (all in minutes), as registered in the central OR Benchmark database

Total Procedure Time Anesthesia-Controlled Time

Surgeon-Controlled Time

UMC N Mean SD Median Mean SD Median Mean SD Median

UMC1 34,316 160 109 131 34 18 30 126 101 100

UMC2 52,329 181 126 142 43 24 36 138 112 104

UMC3 70,264 178 123 146 44 24 39 134 110 104

UMC4 41,266 152 121 113 32 17 27 120 112 84

UMC5 45,955 162 120 130 36 20 31 126 108 96

UMC6 86,128 127 104 92 30 20 26 97 92 65

Total 330,258 158 119 124 37 22 31 121 106 90

Figure 2A shows a scatter plot depicting the aTPT and pTPT (current prediction) in minutes of all observations at all six UMCs. The estimated regression line of the current prediction method is: aTPT = 10.94 + 1.06*pTPT. For example, if TPT was predicted to take 140 minutes, aTPT had a duration of 159 minutes. TPT was underestimated by 19 minutes.

The analysis of the empirical data from six UMCs (N=330,258) demonstrated a mean (SD) ratio aSCT/aACT of 3.61 (2.91) and a median of 2.90. These results indicate that aSCT was approximately three times greater than aACT. The mean of 3.61 was rounded down to 3, to correct for the influence of data outliers, because the variables used in this study were not normally distributed. Hence, on an overall level: (aACT/aTPT) x 100 = 25% and (aACT/aSCT) x 100 = 33%. Based on these results, rpACT and rpTPT were computed as follows:

rpACT = pSCT x 0.33rpTPT = pSCT + (SCT x 0.33) orrpTPT = pSCT x 1.33

See also Table 1 in which the conducted calculations are illustrated.

Figure 3A shows a scatter plot depicting the aTPT and rpTPT in minutes of all observations at all six UMCs. The estimated regression line of this revised prediction strategy is: aTPT = 23.3 + 0.83*rpTPT. For example, if TPT was predicted to take 140 minutes, aTPT had a duration of 139.5 minutes. TPT was overestimated by 0.5 minutes.

Figures 2B and 3B both illustrate the prediction errors and predicted values in minutes. These plots show that the prediction errors increased rapidly in the current situation and varied with a wider range than in the revised prediction strategy. This is also emphasized by Figure 4 showing stacked histograms of the prediction errors for both prediction methods. In the current situation prediction errors ranged from -688 minutes to +488 minutes; in the revised strategy prediction errors ranged from -219 minutes to +231 minutes. Figure 4 merely shows the prediction errors for both methods on the scale from -300 minutes to +300 minutes.

71

Ane

sthe

sia-

cont

rolle

d Ti

me

and

Ope

ratin

g R

oom

Sch

edul

ing

Figure 2. (A) Current prediction method (fixed ACT): scatter plot of actual vs predicted total procedure time in minutes. ACT = anesthesia-controlled time

(B) Current prediction method (fixed ACT): scatter plot of prediction error (actual - predicted total procedure time [TPT]) vs predicted total procedure time in minutes. ACT = anesthesia-controlled time

Figure 3. (A) Revised prediction strategy (relative ACT): scatter plot of actual vs revised predicted total procedure time in minutes.ACT = anesthesia-controlled time

(B) Revised prediction strategy (relative ACT): scatter plot of prediction error (actual – revised predicted total procedure time [TPT]) vs revised predicted TPT in minutes.ACT = anesthesia-controlled time

aTPT=10.94+1.06*pTPT

aTPT=23.3+0.83*rpTPT

72

3

Table 3 shows the mean (SD) ratio aSCT/aACT and median ratio aSCT/aACT differentiated per surgical department using the data of all six UMCs. Cardiothoracic Surgery and Neurosurgery, both characterized by significantly longer aSCT than other surgical departments, demonstrated a ratio aSCT/aACT of four. Orthopedic Surgery and Plastic Surgery also showed a ratio of four, indicating that the recommended scheduling rule ‘rpTPT = pSCT x 1.33’ is preferably differentiated per surgical department, e.g. adjusted to ‘rpTPT = pSCT x 1.25’ for Cardiothoracic-, Neuro-, Orthopedic- and Plastic Surgery.

Figure 4. Stacked histograms of the prediction errors (actual - predicted) in minutes for both prediction methods

Current prediction method

Prediction error (minutes)

actual - predicted

Perc

en

t

Revised prediction method

73

Ane

sthe

sia-

cont

rolle

d Ti

me

and

Ope

ratin

g R

oom

Sch

edul

ing

Tabl

e 3.

Des

crip

tive

stat

istic

s of

act

ual T

otal

Pro

cedu

re T

ime,

act

ual A

nest

hesi

a-C

ontro

lled

Tim

e an

d ac

tual

Sur

geon

-Con

trolle

d Ti

me

(all

in m

inut

es),

as

regi

ster

ed in

the

cent

ral O

R B

ench

mar

k da

taba

se, a

nd th

e ra

tio o

f act

ual S

CT/

AC

T di

ffere

ntia

ted

per s

urgi

cal d

epar

tmen

t usi

ng th

e da

ta o

f all

six

UM

Cs

Act

ual T

otal

Pro

cedu

re

Tim

eA

nest

hesi

a-C

ontr

olle

d Ti

me

Surg

eon-

Con

trol

led

Tim

eR

atio

SC

T/A

CT

NM

ean

SDM

edia

nM

ean

SDM

edia

nM

ean

SDM

edia

nM

ean

SDM

edia

n

Car

diot

hora

cic

Sur

gery

29,4

0826

411

526

159

2556

205

106

201

3.98

2.48

3.49

Gen

eral

Sur

gery

76,2

0317

312

014

340

2434

133

106

106

3.61

2.84

2.97

Ear

-Nos

e-Th

roat

Sur

gery

41,5

5112

911

393

3116

2998

105

613.

163.

032.

18

Ora

l & M

axill

ofac

ial S

urge

ry13

,170

165

130

130

3818

3512

712

194

3.45

2.79

2.76

Neu

rosu

rger

y23

,969

216

143

170

4524

4017

113

212

84.

193.

353.

30

Oph

thal

mol

ogy

36,0

8677

4169

2112

1956

3549

3.54

2.93

2.75

Orth

oped

ic S

urge

ry35

,184

148

8613

435

2031

112

7710

03.

732.

903.

03

Pla

stic

Sur

gery

24,0

0114

812

711

232

1928

116

118

823.

973.

323.

03

Uro

logy

27,2

1013

410

199

3217

2810

292

703.

262.

662.

51

Obs

tetri

cs &

Gyn

aeco

logy

23,4

7613

892

113

3317

2910

583

823.

432.

672.

78

Tota

l33

0,25

815

811

912

437

2231

121

106

903.

612.

912.

90

74

3

DISCUSSION

This Dutch, nationwide, multicenter study shows that in a university hospital environment, grossing up SCT by 33% to account for rpACT, can improve the prediction of TPT if this methodology is adopted. This confirms that employing a fixed time period for ACT (e.g. 20 minutes), is unsuitable because like SCT, ACT is subject to variability. The results affirm that ACT is a considerable part of TPT, which should be scheduled just as realistically as SCT. Robust OR schedules need to anticipate SCT as well as ACT. ACT should be predicted apart from SCT, as a separate time period instead of one predicted time period for TPT.

A previous study by Overdyk et al. in 1998, from the Medical University of South Carolina, a 550-bed teaching hospital and tertiary referral center (N=1,881 cases), found that: “surgeon-controlled time is approximately four and a half times greater than anesthesia-controlled time”19. These authors suggested grossing up SCT by 22% to account for ACT19.

Our study based on an extensive database collected in a considerable research setting of six centers, shows that this proportion should be higher and at least 33% (rpTPT = pSCT x 1.33). More accurate prediction rules may lead to less over-utilized OR time and reducing the number of case cancellations2,5.

Potential sources for the difference among Overdyk’s study results19 and these recent results are the size of the hospitals and the number of cases investigated. All UMCs have a bed capacity that varies between 715 and 1,339 beds, and our study investigated 330,258 cases. Both considerably larger than previous research.

Figures 2A, 2B and 4 demonstrate that the current prediction method frequently underestimated aTPT, leading to the risk of OR days running late and case cancellations. Figures 3A, 3B and 4 demonstrate that employing our recommended scheduling rule results in lower prediction errors and especially less underestimation of aTPT. Moreover, the increasing size of the prediction errors with increasing predicted values suggests the use of a proportion of time rather than a fixed number of minutes methodology for ACT.

In many hospitals, surgeons make a routine prediction of the procedure time needed. Though it has been shown that surgeons tend to underestimate the time needed to perform a procedure20-22. In some facilities, historical procedure times are taken as a reference to predict the duration of future cases. This methodology has been recommended by earlier studies14,23-27. Some centers employ an approximation of pACT in the pTPT9. Often as a fixed number of minutes. All types of anesthesia, such as general, regional, local, combined general with epidural and monitored anesthesia care, were included in the study. The participating UMCs do not use anesthetic induction rooms distinct from the operating room. In these Dutch centers all activities related to anesthesia care are provided in the OR and therefore influence OR utilization. There is, however, one center in which occasionally the insertion of the epidural catheter is placed in the preoperative holding area, but this is not always possible due to

75

Ane

sthe

sia-

cont

rolle

d Ti

me

and

Ope

ratin

g R

oom

Sch

edul

ing

the large amount of patients. Referring to the employment of a fixed number of minutes for pACT, Escobar et al.28 found significant variation in anesthesia release time and concluded that for OR scheduling purposes assigning a constant fixed time for anesthetic induction is inappropriate. This recent study produced results which corroborate that conclusion.

There are, however, valid limitations to this study. Although the revised prediction strategy has decreased the prediction errors – see Figure 4 – it is obvious from Figure 3A that for any given predicted total procedure time there is still substantial variability in the actual total procedure time. Scheduling surgical procedures will remain a multifactorial and therefore complex process.

Because data were gathered in tertiary referral centers only, general applicability of the findings may be restricted. The mean (SD) aTPT of 158 (119) minutes and the median of 124 minutes in the present study reflect that the complexity of procedures is potentially greater than in other facilities. The application of this scheduling rule in non-academic facilities has yet to be studied.

Another issue with this and past studies related to OR scheduling and OR efficiency lies in the way data are collected19. In the Netherlands, the OR departments of all eight UMCs established a benchmarking collaboration in 2004, continuing to this day. Each UMC submits the data records of all surgical cases performed to a central OR Benchmark database. All data are prospectively, electronically entered in real-time by the OR nursing staff into a Hospital Information System per UMC and subsequently confirmed by the surgeon and anesthesiologist in charge. The individual databases of each of the eight UMCs are originally intended for administrative and managerial purposes. We acknowledge the potential, virtually unavoidable biases stemming from this data collection source (administrative/nursing database) and agree with Overdyk’s19 remark that it might even be impossible to exclude bias when data collection depends on human individuals instead of automatic electronic time recording systems.

The estimated magnitude of this ‘human bias’ in our longitudinal study is considered to have a small impact because of the long-term stable nature of data capture. It involves repeated and continuous measurement of the same parameters over a long period of time. In this respect, we have assessed the OR benchmark data and found that parameters over all these years (2005-2011) either show a consistent picture over the years, a gradual increase or a gradual decrease. Furthermore, the differences between the UMCs also show a consistent picture, which does not indicate that human bias is of imperative size.

Scheduling surgical procedures is complex because a procedure entails several elements subject to variability, such as room setup and takedown, patient positioning, prepping and draping, as well as the two principal components SCT and ACT. More variability factors such as OR team member characteristics and their experience (attending, fellow, resident, trainee and experience in years) can be of influence on aTPT. Even for experienced anesthesiologists, it is often difficult to predict how long the anesthetic induction for a specific surgical case will

76

3

take9. Factors such as ASA physical status, age, anesthetic technique (e.g. monitors, lines, pain management procedures), working with trainees and residents in a teaching setting and surgical procedure have shown to affect aACT and aTPT significantly3,9,8,29. Because the central OR Benchmark database was not designed to register all of these variability factors, this study could not investigate their impact on aTPT. We believe, however, that these factors are most likely equally distributed among the different UMCs and therefore are of limited impact on our conclusions.

Future studies, which take more variability factors into account, will need to be undertaken. On individual hospital level these factors are partially available. It would also be interesting to compare aSCT among surgeons regarding the same procedure, as well as aACT among anesthesiologists regarding the same anesthetic technique. Using historical times per surgeon and per procedure to schedule pSCT is not new14,23-27, using historical times per anesthesiologist, however, is not common in the Netherlands. Recently one Dutch UMC adopted a system of scheduling pACT based on historical times per anesthesiologist and per anesthetic technique. The implementation of this process started at the end of 2012 and further research is needed to assess the value and effects of this methodology in practice.

If readers wish to implement the suggested scheduling rule (rpTPT = SCT x 1.33), it is recommended to calculate the relevant proportion aSCT/aACT using their own specific historical procedure data. Additionally, it is endorsed to refine the scheduling rule per surgical department, which was indicated by the differentiation of the ratio in the results section (Table 2).

Efficient OR management demands the accurate prediction of the times needed for all components of care (including the two main elements SCT and ACT) for each surgical procedure3,9,14,30. Supported by an extensive dataset from six UMCs, we advise grossing up the pSCT by 33% to account for pACT (rpTPT = pSCT x 1.33), rather than employing a fixed number of minutes methodology. Thirty-three percent is a higher proportion than reported in earlier research19. This recommendation will improve OR scheduling, which might result in the reduction of over-utilized OR time and case cancellations, and therefore in more efficient use of limited OR resources2,5,11.

77

Ane

sthe

sia-

cont

rolle

d Ti

me

and

Ope

ratin

g R

oom

Sch

edul

ing

REFERENCES


2. Cardoen B, Demeulemeester E, Belien J. Operating room planning and scheduling: A literature review. European Journal of Operational Research 2010;201:921-32.




6. Agnoletti V, Buccioli M, Padovani E, et al. Operating room data management: improving efficiency and safety in a surgical block. BMC Surg 2013;13:1-11.

7. Devi SP, Rao KS, Sangeetha SS. Prediction of surgery times and scheduling of operation theaters in ophthalmology department. J Med Syst 2012;36:415-30.

8. Dexter F, Dexter EU, Masursky D, Nussmeier NA. Systematic review of general thoracic surgery articles to identify predictors of operating room case durations. Anesth Analg 2008;106:1232-41.


10. Gillespie BM, Chaboyer W, Fairweather N. Factors that influence the expected length of operation: results of a prospective study. BMJ Qual Saf 2012;21:3-12.

11. Lehtonen JM, Torkki P, Peltokorpi A, Moilanen T. Increasing operating room productivity by duration categories and a newsvendor model. Int J Health Care Qual Assur 2013;26:80-92.

12. Pandit JJ, Tavare A. Using mean duration and variation of procedure times to plan a list of surgical operations to fit into the scheduled list time. European Journal of Anaesthesiology 2011;28:493–501.

13. Smith CD, Spackman T, Brommer K, et al. Re-engineering the operating room using variability methodology to improve health care value. J Am Coll Surg 2013;216:559-68.


15. Fone D, Hollinghurst S, Temple M, et al. Systematic review of the use and value of computer simulation modelling in population health and health care delivery. J Public Health Med 2003;25:325-35.

16. Kazemier G, van Veen-Berkx E. Comment on “Identification and use of operating room efficiency indicators: the problem of definition”. Can J Surg 2013;56:E103-4.

17. Van Veen-Berkx E, Elkhuizen SG, Kalkman CJ, Buhre WF, Kazemier G. Successful Interventions to Reduce First-Case Tardiness in Dutch University Medical Centers. Results of a Nationwide Operating Room Benchmark Study. Am J Surg 2013;In press.

18. Dexter F, Coffin S, Tinker JH. Decreases in anesthesia-controlled time cannot permit one additional surgical operation to be reliably scheduled during the workday. Anesth Analg 1995;81:1263-8.


78

3

20. Alvarez R, Bowry R, Carter M. Prediction of the time to complete a series of surgical cases to avoid cardiac operating room overutilization. Can J Anaesth 2010;57:973-9.


22. Dexter F, Wachtel RE, Epstein RH, McIntosh C, O’Neill L. Allocative efficiency vs technical efficiency in operating room management. Anaesthesia 2007;62:1290-1; author reply 1-2.

23. Pandit JJ, Carey A. Estimating the duration of common elective operations: implications for operating list management. Anaesthesia 2006;61:768-76.

24. Zhou J, Dexter F, Macario A, Lubarsky DA. Relying solely on historical surgical times to estimate accurately future surgical times is unlikely to reduce the average length of time cases finish late. J Clin Anesth 1999;11:601-5.

25. Dexter F, Ledolter J, Tiwari V, Epstein RH. Value of a scheduled duration quantified in terms of equivalent numbers of historical cases. Anesth Analg 2013;117:205-10.

26. Dexter F, Ledolter J. Bayesian prediction bounds and comparisons of operating room times even for procedures with few or no historic data. Anesthesiology 2005;103:1259-167.

27. Dexter F, Traub RD, Fleisher LA, Rock P. What sample sizes are required for pooling surgical case durations among facilities to decrease the incidence of procedures with little historical data? Anesthesiology 2002;96:1230-6.

28. Escobar A, Davis EA, Ehrenwerth J, et al. Task analysis of the preincision surgical period: an independent observer-based study of 1558 cases. Anesth Analg 2006;103:922-7.

29. Urman RD, Sarin P, Mitani A, Philip B, Eappen S. Presence of anesthesia resident trainees in day surgery unit has mixed effects on operating room efficiency measures. Ochsner J;12:25-9.

30. Dexter F, Macario A. Applications of information systems to operating room scheduling. Anesthesiology 1996;85:1232-4.

79

Ane

sthe

sia-

cont

rolle

d Ti

me

and

Ope

ratin

g R

oom

Sch

edul

ing

4Effect of Individual Surgeons

and Anesthesiologists

on Operating Room Time

Ruben P.A. van Eijk, MDElizabeth van Veen-Berkx, MScGeert Kazemier, MD, PhDMarinus J.C. Eijkemans, PhD

Anesthesia&Analgesia.2016.123:445-51

82

4

Abstract

Background. Variability in operating room (OR) time causes over- and underutilization of the available operating rooms. There is evidence that for a given type of procedure, the surgeon is the major source of variability in OR time. The primary aim was to quantify the variability between surgeons and anesthesiologists. As illustration, the value of modeling the individual surgeons and anesthesiologist for OR time prediction was estimated.

Methods. Operating room data containing 16,480 cases originated from a general surgery department. The total amount of variability in OR time accounted for by the type of procedure, first and second surgeon and the anaesthesiologist was determined using linear mixed models. The effect on OR time prediction was evaluated as reduction in overtime and idle time per case.

Results. Differences between first surgeons can account for only 2.9% (2.0 – 4.2) of the variability in OR time. Differences between anesthesiologists can account for 0.1% (0.0 - 0.3) of the variability in OR time. Incorporating the individual surgeons and anesthesiologists led to an average reduction of overtime and idle time of 1.8 (CI: 1.7 – 2.0, 10.5% reduction) minutes and 3.0 (CI: 2.8 – 3.2, 17.0% reduction) minutes, respectively.

Conclusions. In comparison with the type of procedure, differences between surgeons account for a small part of OR time variability. The impact of differences between anaesthesiologists on OR time is negligible. A prediction model incorporating the individual surgeons and anaesthesiologists has an increased precision but improvements are likely too marginal to have practical consequences for OR scheduling.

83

Effe

ct o

f Sur

geon

s an

d A

nest

hesi

olog

ists

on

Ope

ratin

g R

oom

Tim

e

INTRODUCTION

In times with rising healthcare costs and scarcity in healthcare budgets, efficiency within hospitals is becoming of the utmost importance. Over 60% of the patients admitted to a hospital are treated in the operating room (OR). The OR is therefore one of the major consumers of the total healthcare budget (1). Accurate OR scheduling is critical for efficiency but remains a challenging problem due to the high inherit variability in OR time (2; 3). Variability in OR time causes over- and underutilization of the available operating room capacity. This inefficient use and overutilization of OR time lead not only to a waste of resources, but may also lead to dissatisfaction and reduced motivation of the surgical staff and increased patient waiting times (4).

In the literature there is a wide availability of methods to estimate OR time ranging from simple estimations based on the sample mean to complex Bayesian prediction models (5). In general, the accuracy of prediction models for continuous outcomes (such as OR time) depends strongly on its capability to model the variability in an outcome (6; 7). To increase the accuracy of OR time prediction models, it is therefore critical to identify sources of variability in OR time. Subsequently, future prediction models can then be expanded to account for these sources of variability.

There are several research papers that investigated sources of variability in OR time (1,8,10). Not surprising, the type of procedure is the single most important source of variation (1, 4,10). The effect of patients characteristics (i.e. age, gender and BMI) on OR time is rather small in comparison with the type of procedure (1, 11). Sources of variability in anesthesia time, and thus total OR time, were identified as well and are mainly due to the anesthetic technique and type of procedure (2; 12,13). Several studies showed the importance of surgical team characteristics on OR time (1,9,10,14-18). It is believed that after the type of procedure, the surgeon is the single most important source of variability in OR time (10). The variability in OR time between surgeons can be explained by a difference in work rate (9,10); however, it seems that not only a difference in work rate but also that the surgeon’s age and experience with a certain procedure have important effects on OR time (1,9,16). In particular the age of the youngest and oldest surgeon are important during a certain procedure; they may act as surrogates for the surgeon’s level of experience and the difficulty of a certain procedure (1). Furthermore, the accuracy of the surgeon’s estimate of the OR time may differ between surgeons and may thus lead to an increased variability in OR time (18). All these factors are properties of the individual surgeon or properties of combinations of surgeons. It seems therefore plausible that the individual surgeon should account for a fairly large part of the variability in OR time. The effects of these specific surgeon characteristics on OR time are well shown in the literature, however, a clear quantification of the exact amount of variability that can be accounted to the individual surgeons is currently lacking.

84

4

Several studies have reported prediction models using the specific combination of the surgeon and type of procedure and have shown to improve the accuracy of OR time estimation (19). However, these models do not take into account the individual characteristics of the patient. As was shown before, modeling individual case characteristics substantially improved the prediction of OR time (1). Thus, a prediction model based on individual case characteristics and accounting for the variability between surgeons might thus further improve the prediction of OR time; however, as was stated previously, currently, there is no clear quantification available of the surgeon’s variability in OR time, and the exact effects of individual surgeons on the prediction of OR time are unknown. Therefore, the primary aim of this study is to quantify the variability between surgeons in comparison with the type of procedure. Because anesthesia time is an important part of the total OR time (12), our secondary aim was to quantify the variability between anesthesiologists. As illustration, the value of modeling these sources variability for the prediction of OR time will be estimated.

MATERIALS AND METHODS

Data and SubjectsTo reliably estimate the additional value of modeling the individual surgeons and anesthesiologists during a procedure, we used the same dataset as from our previous model based on individual case characteristics (1). In short, data of the operative sessions originated from the Erasmus University Medical Center, Department of General Surgery (Rotterdam, the Netherlands). All operative sessions were registered electronically since January 1993. From this date until June 2005, all consecutive elective operations performed were included in our dataset. Data were matched with data from the general electronic hospital information system for details about risk factors of surgical complications (i.e. cardiovascular diseases or diabetes). The final database contained 17,412 cases, classified into 253 different types of procedures according to the main procedure during a session. When multiple procedures were performed during a case (i.e. breast reconstruction after mastectomy), the operation was coded according to the main procedure. The main procedure was determined from a priority list that was constructed by surgeons of the general surgery department. This method was preferred over statistical determination of the longest procedure (20), because some procedures were never performed in isolation. For accurate determination of the variability between surgeons and anesthesiologists, cases without an anesthesiologist and/or a second surgeon were excluded. The second surgeon is defined as the first registered assistant surgeon during a procedure. After removal, 16,389 cases, classified into 251 different surgical procedures, remained for data analysis.

The total operating room time was defined as the elapsed time between the patients’ arrival at and departure from the OR. The recorded data contained: (1) procedure

85

Effe

ct o

f Sur

geon

s an

d A

nest

hesi

olog

ists

on

Ope

ratin

g R

oom

Tim

e

characteristics (expected duration, number of separate procedures, laparoscopic or open surgery), (2) operating team characteristics (surgeon, second surgeon, anesthesiologist, total of the ages as a measure of the combined experience, age of the youngest and oldest surgeon and number of surgeons and anesthesiologists) and (3) patient characteristics (age, sex, the number of admissions before the operation, length of current admission, BMI and cardiovascular risk factors (diabetes, hypercholesterolemia, hypertension, heart failure, cerebrovascular accident, COPD, renal and cardiac diseases)). Log transforming the OR time was necessary due to right skewness (1,10,20). A more detailed description of multiple imputation of missing data and the corrections that were necessary to make the data suitable for prediction modeling can be found in our previous article (1).

Statistical AnalysisThe primary aim was to estimate how much of the total amount of variability in log OR time can be accounted for by the type of procedure, surgeon and anesthesiologist. First, the total amount of variability (or variance) in log OR time was estimated. Subsequently, for the type of procedure, surgeon or anesthesiologist their unique part of this total variance was determined. The random effects part of linear mixed models (LMM) was used to estimate these variance components, as LMM allow inclusion of infrequent procedures, surgeons or anesthesiologists, even those that only occurred once (1,5).

It was hypothesized that for certain procedures the difference between surgeons (or anesthesiologists) could be bigger than for other procedures. Thus, for highly technical procedures, surgeons may perform less similar than for relatively easy routine procedures. This means that variability between surgeons and anesthesiologists depends on the type of procedure. Furthermore, the difference between first surgeons could depend on the second surgeon (and vice versa). In example, the OR time for a certain procedure performed by two senior surgeons is probably different from the OR time when there is one senior and one junior surgeon. Therefore, interaction terms were constructed for the type of procedure with the first surgeon, second surgeon and anesthesiologist and for the first surgeon with the second surgeon. To illustrate the construction of the final multivariate random effect LMM with only an intercept, the equation was given by:

Yij = β0 + μ0 procedure + μ0 surgeon I + μ0 surgeon II + μ0 anesthesiologist + μ0 procedure*surgeon I + μ0

procedure * surgeon II + μ0 procedure*anesthesiologist + μ0 surgeon I*surgeon II + εij

In this model, Yij represents the predicted log OR time, β0 the intercept and μ0 … the various random effects. εij is the residual variance and is the part that the model cannot explain. The total variance in this model is the sum of all random effect variances plus the residual variance. Dividing the variance attributed to a certain random effect (i.e. the variance of

86

4

μ0 procedure) by the total variance, will give the percentage of the total variance that can be accounted for by the random effect (known as Intraclass Correlation Coefficient (ICC)(21)). The ICC thus represents how much of the total variability in log OR time can be accounted for by the surgeons, anesthesiologists and types of procedures.

The second aim of the analysis was to illustrate the improvement of a prediction model incorporating the aforementioned random effects. All models were corrected for predictive factors as fixed effects (surgeon’s estimate and the operation, team and patient characteristics). The total variance was calculated in a base model with only a random intercept for the type of procedure. The random effects terms were subsequently added to the model to evaluate how much the unexplained variance could be reduced. The ratio explained/unexplained variance was given by the adjusted R2. In example, when the adjusted R2 was 70%, this indicated that 70% of the variance in OR time could be explained by the model and 30% was left unexplained. To quantify the improvement of a new model compared to the base model, the gain in adjusted R2 was calculated as (R2

model - R2 base) / (1 - R2

base). Model fit was further evaluated by Akaike Information Criteria (AIC) and likelihood ratio tests (LRT). For each random effect, the absolute reduction in over and idle time (in minutes) was estimated. Confidence intervals around these estimates were bootstrapped by drawing 2000 random samples. Predictions in log OR time were multiplied by a smearing factor to reduce back-transformation bias (22). After fitting the random effects parts, the fixed effect part of the model was stepwise reduced based on the AIC to obtain the most parsimonious model. For the final model, the ICCs for the type of procedure, surgeons and anesthesiologist were recalculated. The stability of the variance estimates and their 95% confidence intervals were evaluated by refitting the model several times and determining ζ (zeta) for each random effect (23). LMM were fitted using the lmer function in the R package lme4 (version 1.1-11) (31).

RESULTS

The final database contained 251 different types of procedures, 215 first surgeons, 243 second surgeons and 168 anesthesiologists. Figure 1 shows the median OR time for all types of procedures, surgeons and anesthesiologists univariately. The type of procedure has the widest interquartile range and showed the highest variability in median OR time. Table 1 summarizes the ICCs obtained from univariate models (incorporating only one random effect) with and without correction for the predictive factors. In accordance with figure 1, the type of procedure caused the largest variability in log OR time and accounted for 33.1% (CI: 27.0 – 41.0) of the total variability in log OR time in a corrected model. The overall effect of differences between the first surgeons on variability in OR time was small (ICC 3.9% (CI: 2.8 – 5.5)). However, when the type of procedure was considered, differences between first

87

Effe

ct o

f Sur

geon

s an

d A

nest

hesi

olog

ists

on

Ope

ratin

g R

oom

Tim

e

0

200

400

600

Type of Procedure Surgeon I Surgeon II Anesthesiologist

Med

ian

OR

tim

e (m

inut

es)

Univariate distribution of OR time

Figure 1. Univariate distribution of OR time. Distribution of median OR time according to the type of procedure, surgeons and anesthesiologists. The type of procedure has the largest interquartile range (IQR) in OR time. The tails of the boxplot for all grouping variables are unequal in length, indicating right skewness in OR time. Grouping the dataset by anesthesiologists shows the smallest IQR in OR time.

Table 1. Univariate assessment of the various random effects with and without correction for predictive factors expressed as ICC

Random effect ICC Without correction

(%) ICC With correction

(%)

Procedure 71.1 (CI: 59.6 – 85.8) 33.1 (CI: 27.0 – 41.0)

Surgeon I 27.8 (CI: 21.5 – 36.2) 3.9 (CI: 2.8 – 5.5)

Surgeon II 7.8 (CI: 5.6 – 10.7) 2.3 (CI: 1.5 – 3.5)

Surgeon I * Surgeon II 35.1 (CI: 32.5 – 37.9) 7.1 (CI: 5.7 – 8.6)

Anesthesiologist 2.8 (CI: 1.8 – 4.5) 0.1 (CI: 0.0 – 0.1)

Surgeon I * Procedure 76.0 (CI: 72.6 – 79.6) 31.4 (CI: 28.5 – 34.5)

Surgeon II * Procedure 77.3 (CI: 73.9 – 80.8) 30.2 (CI: 27.2 – 33.3)

Anesthesiologist * Procedure 73.9 (CI: 70.6 – 77.4) 22.2 (CI: 19.6 – 25.0)

The model was fitted univariately, all models contain only one random effect (i.e. for type of procedure the mixed model equation was given by: Yij = β0 + μ0 procedure + (β1 – βn predictive factors) + εij). ICC = intraclass correlation coefficient = (variance of the random effect / total variance in model) * 100%. CI = 95% confidence interval. Correction predictive factors: (1) procedure characteristics (expected duration, number of separate procedures, laparoscopic or open surgery), (2) operating team characteristics (total age of the surgeons as a measure of the combined experience, age of the youngest and oldest surgeon and number of surgeons and anesthesiologists) and (3) patient characteristics (age, sex, the number of admissions before the operation, length of current admission, BMI and cardiovascular risk factors (hypercholesterolemia and hypertension)).

88

4

surgeons accounted for 31.4% (CI: 28.5% – 34.5%) of the total variability in log OR time in a univariate corrected model.

Table 2a shows the model in which the various random effects were subsequently added to the base model. The base model consisted of the type of procedure as random effect and all predictive factors. The base model was identical to our earlier work, but exclusion of data with only one surgeon led to an altered adjusted R2 of 79.5 instead of the 79.8 reported previously (1). The final R2 relatively increased with 17.6% (CI: 15.1% – 20.0%) when incorporating all significant random effects. In other words, incorporating random effects for surgeons and anesthesiologists can explain 17.6% of the previously unexplained variance. Both interaction terms for the first and second surgeons with the type of procedure were significant additions (p < 0.001). This indicates that the differences between first and second surgeons depend on the type of procedure. Thus, for some procedures the surgeons will perform more similar than for other procedures. Furthermore, the difference between first surgeons depends also on the second surgeon (p < 0.001). Table 2b shows the effect on the accuracy of a prediction model in over and idle time. When the final significant interaction was added to the model (procedure * surgeon II, p < 0.001), this led to an average reduction of over and idle time of 1.9 (CI: 1.8 – 2.0) minutes and 3.1 (CI: 3.0 – 3.3) minutes per case, respectively.

After inclusion of all significant random effects from table 2a, the fixed part of the model with the predictive factors was stepwise reduced based on the AIC. The excluded predictive factors are given in table 3. Interestingly, most of the excluded factors were related to either characteristics of the surgeon or anesthesiologist (i.e. age of the oldest and the youngest surgeon during a procedure) or related to the kind of procedure the surgeon performs (i.e. length of current admission).

The random effects part of the final model is summarized in table 4. Overall, differences between first surgeons can account for only 2.9% (2.0 – 4.2) of the variability in log OR time. Differences between anesthesiologists can account only for 0.1% (0.0 - 0.3) of the variability in log OR time. When the type of procedure is considered, differences between first surgeons can account for 5.5% (4.3 – 6.8) of the variability in log OR time. Figure 2 shows that there is no severe deviation from normality for the random effects, indicating reliable estimations and stability of the model. The final model, incorporating the individual surgeons and anesthesiologists, had an adjusted R2 of 83.1% (relative increase of 17.6%) and led to an average reduction of over time and idle time of 1.8 (CI: 1.7 – 2.0, 10.5% reduction) minutes and 3.0 (CI: 2.8 – 3.2, 17.0% reduction) minutes, respectively. Although this is a significant reduction (p < 0.001) in over and idle time, the differences are likely too marginal to have practical consequences for OR scheduling.

89

Effe

ct o

f Sur

geon

s an

d A

nest

hesi

olog

ists

on

Ope

ratin

g R

oom

Tim

e

Table 2a. Base model extended with random effects for the surgeons and anesthesiologists: effect on residual variance

Parameter AIC Adjusted R2

(%)Adjusted R2 gain

(%)p-value

Base model 4006.4 79.5 - -

+ Surgeon I 3685.9 80.1 2.9 (CI: 1.0 – 5.4) P < .001

+ Surgeon II 3464.2 80.6 5.4 (CI: 3.4 – 7.3) P < .001

+ Surgeon I * surgeon II 3435.4 81.0 7.3 (CI: 5.4 – 9.8) P < .001

+ Anesthesiologist 3431.9 81.1 7.8 (CI: 5.4 – 9.8) P = .019

+ Surgeon I * procedure 3264.3 82.4 14.1 (CI: 11.7 – 16.1) P < .001

+ Surgeon II * procedure 3199.9 83.1 17.6 (CI: 15.1 – 20.0) P < .001

+ Anesthesiologist * procedure 3201.2 83.2 18.0 (CI: 15.1 – 20.5) P = .419

Base model = random effect for procedure + all predictive factors. Predictive factors: (1) procedure characteristics (expected duration, number of separate procedures, laparoscopic or open surgery), (2) operating team characteristics (total age of the surgeons as a measure of the combined experience, age of the youngest and oldest surgeon and number of surgeons and anesthesiologists) and (3) patient characteristics (age, sex, the number of admissions before the operation, length of current admission, BMI and cardiovascular risk factors (hypercholesterolemia and hypertension)). AIC = Akaike Information criteria. P-values are based on the incremental gain as compared to the preceding model in the table.

Table 2b. Base model extended with random effects for the surgeons and anesthesiologists: effect on overtime and idle time in minutes

Parameter Overtime(Minutes)

Gain over

(Minutes)Idle time(Minutes)

Gain idle

(Minutes)

Base model 17.2(CI: 16.7 – 17.8)

- 17.6(CI: 17.2 – 18.8)

-

+ Surgeon I 16.8(CI: 16.2 – 17.3)

0.4(CI: 0.4 – 0.5)

17.1(CI: 16.6 – 17.5)

0.5(CI: 0.5 – 0.7)

+ Surgeon II 16.6(CI: 16.1 – 17.1)

0.6(CI: 0.5 – 0.7)

16.9(CI: 16.6 – 17.3)

0.7(CI: 0.6 – 0.8)

+ Surgeon I * surgeon II 16.4(CI: 15.9 – 16.9)

0.8(CI: 0.7 – 0.9)

16.4(CI: 16.0 – 16.8)

1.2(CI: 1.1 – 1.4)

+ Anesthesiologist 16.4(CI: 15.9 – 17.0)

0.8(CI: 0.7 – 0.9)

16.4(CI: 16.0 – 16.8)

1.2(CI: 1.1 – 1.4)

+ Surgeon I * procedure 15.7(CI: 15.2 – 16.2)

1.5(CI: 1.4 – 1.7)

15.2(CI: 14.8 – 15.5)

2.4(CI: 2.3 – 2.6)

+ Surgeon II * procedure 15.3(CI: 14.8 – 15.8)

1.9(CI: 1.8 – 2.0)

14.5(CI: 14.2 – 14.8)

3.1(CI: 3.0 – 3.3)

+ Anesthesiologist * procedure 15.2(CI: 14.7 – 15.7)

2.0(CI: 1.9 – 2.2)

14.5(CI: 14.2 – 14.9)

3.1(CI: 2.9 – 3.2)

Overtime and idle time: mean overtime and idle time per case, calculated as (predicted OR time – observed OR time)/amount of cases. The gain was calculated as (mean over- or idle time model 1) – (mean over- or idle time model 2). CI = 95% confidence interval, calculated by bootstrapping using 2000 random samples.

90

4

Table 3. Reducing the fixed effects part based on Akaike Information Criteria (AIC)

Parameter AIC

Base model 3199.9

- Age of oldest surgeon * procedure 3146.3

- Number of anesthesiologists 3142.3

- Age of patient 3139.6

- Number of preceding operations 3137.4

- Age of oldest anesthesiologist 3135.4

- Age of youngest surgeon 3133.8

- Hypertension 3132.1

- Number of admissions before the operation * procedure 3131.6

- Length of current admission * procedure 3127.4

- Length of current admission 3125.4

- Year 3125.8

Base model = all significant random effects from table 2a + all predictive factors. For each predictive factor the loss or gain in AIC was calculated if it was dropped from the model. The predictive factor with the lowest AIC when it was excluded from the model, was definitely excluded. This process was repeated until no relevant AIC reduction could be achieved. If the AIC was less than 2 points different, the more parsimonious model was chosen.

Table 4. Multivariate ICC for the random effects in the final model

Parameters ICC 95% Confidence Interval

Procedure 31.6 25.6 – 39.4

Surgeon I 2.9 2.0 – 4.2

Surgeon II 1.9 1.2 – 2.9

Surgeon I * surgeon II 0.8 0.3 – 1.5

Anesthesiologist 0.1 0.0 – 0.3

Surgeon I * Procedure 5.5 4.3 – 6.8

Surgeon II * Procedure 4.0 2.9 – 5.4

Multivariate random effects model corrected for predictive factors, the mixed model equation was given by: Yij = β0 + μ0 procedure + μ0 surgeon I + μ0 surgeon II + μ0 anesthesiologist + μ0 procedure*surgeon I + μ0 procedure * surgeon II + μ0 surgeon I*surgeon

II + (β1 – βn predictive factors) + εij. ICC = intraclass correlation coefficient = (variance of the random effect / total variance in model) * 100%. CI = 95% confidence interval.

91

Effe

ct o

f Sur

geon

s an

d A

nest

hesi

olog

ists

on

Ope

ratin

g R

oom

Tim

e

Profile Zeta Plot of the Random Effects

ζ

−2

−1

0

1

2

0.055 0.060 0.065 0.070 0.075 0.080

Surgeon 2 − Procedure interaction

0.065 0.070 0.075 0.080 0.085

Surgeon 1 − Procedure interaction0.015 0.020 0.025 0.030 0.035 0.040

Surgeon 1 − Surgeon 2 interaction

0.16 0.17 0.18 0.19 0.20 0.21

−2

−1

0

1

2

Procedure

−2

−1

0

1

2

0.035 0.040 0.045 0.050 0.055 0.060

Surgeon 2

0.045 0.050 0.055 0.060 0.065 0.070

Surgeon 10.005 0.010 0.015 0.020

Anesthesiologist

0.236 0.238 0.240 0.242 0.244

−2

−1

0

1

2

Residual

Figure 2. Profile Zeta plot of the Random EffectsThe final model was re-estimated multiple times while fixating alternately one parameter and varying the others. This results in slight variations of the random effects. These values are plotted against quantiles of the normal distribution and result in the above graph, which can be interpreted in a similar way as a QQ-plot. The x-axis is in log OR-time. The vertical black lines represent the 60, 80, 90, 95 and 99% confidence intervals. All random effects are approximately on a straight line within their 95% confidence intervals, indicating minimal deviations from normality and reliable estimations of the random effects.

92

4

DISCUSSION

In this study, we identified and quantified the amount of variability in log OR time that can be accounted to differences between the individual surgeons and anesthesiologists. To elaborate on our previous article (1), we evaluated whether individual case characteristics, in this case the individual surgeons and anesthesiologists, can improve the precision of OR prediction. Several studies showed the importance of surgical team characteristics on OR time (1,9,10,14-18). Reported factors are often properties of the individual surgeon (i.e. work rate, age or experience (1,9,10,16)) or properties of combinations of surgeons (i.e. team familiarity and the number of surgeons (1, 17)). Therefore, we hypothesized that modeling the individual surgeons should account for a fairly large part of the total variability in OR time. Surprisingly, differences between first surgeons could only account for 2.9% (CI: 2.1% – 4.2%) of the total variability in log OR time. Differences between anesthesiologists had a negligible effect on OR time and accounted for merely 0.1% (CI: 0.0% – 0.3%). This confirms an earlier report that the anesthesiologist has little impact on OR time (10). The type of procedure accounted for the largest part in variability in OR time (31.6% (CI: 25.6% – 39.4%)). Although the surgeons and anesthesiologists can account for a unique part of the total variability in OR time (relative increase in adjusted R2 of 17.6%), the mean effect on the precision of a prediction model was minimal. Probably the reduction in overutilized OR time by our increase in precision is negligible and has no practical consequences for OR planning (5,24). Furthermore, decision-making would not be affected because a small reduction in overtime rarely changes the decision whether a case is performed or cancelled (25).

The final model contained significant random interaction terms for the first surgeon and second surgeon with the type of procedure. This indicates that the differences between first and second surgeons depend on the type of procedure. Thus, for some procedures the surgeons will perform more similar than for other procedures. Several explanations can be given. First, Strum et al. mentioned that surgeons consistently work in different paces: the work rate effect (10). Differences between surgeons increase proportionately with longer procedures. Thus, for procedures with short median OR time, surgeons will be more similar than for longer procedures in terms of median OR time. Second, several studies have shown that the experience of the surgeon or the surgical team influences the duration of the OR time (17,26,27). These studies illustrate that increased experience (expressed as performance frequency of a procedure) lowers the duration of procedures. Therefore, surgeons with less experience with a certain type of procedure are likely to show more variability in their OR time. At last, these data originate from a university medical center where oncologic procedures were performed by specific surgeons. For oncologic procedures, the discrepancy between procedure times can be high due to incorrect pre-operative tumor staging or conversion of the planned procedures. For example, during laparoscopic tumor resections, conversion rates

93

Effe

ct o

f Sur

geon

s an

d A

nest

hesi

olog

ists

on

Ope

ratin

g R

oom

Tim

e

to open procedures can be as high as 20% (28). Incorrect preoperative staging may reveal inoperable oncology during surgery. In that case, the predicted duration will be much longer than the actual time and this leads to an increased variability in OR time.

This study has several limitations. First of all, the influence of surgeons and anesthesiologists was determined by analyzing the total OR time. The OR time started when the patient entered the OR and ended when the patient was leaving the OR. This may distort the precision of the analysis due to the fact that anesthesia time only consumes 7 – 25 % of the total OR time (4; 12). Because of this disproportionate division of OR time, the effect of the anesthesiologist on OR time is therefore already lower than the surgeon. However, as both the effect of the surgeon and anesthesiologist are small, it is unlikely that the division in surgical and anesthetic time would cause large alterations in our results.

The second limitation of this study is the strict separation between first and second surgeon. For some procedures, the second surgeon of a case could be the first surgeon in another case and vice versa. It is thus not always clear if the first surgeon primarily determined the length of a procedure. Therefore, the variance components we determined separately for first and second surgeons are likely to be related to each other. We partially corrected for this by including a random interaction term for the first and second surgeon. As this term accounted only for 0.8% of the total variability in OR time, it is unlikely that a strict separation generates very different results. A solution to this problem would be to analyze the first and second surgeon as a team. However, that would have prevented us to determine the difference between individual surgeons. Moreover, by analyzing individual surgeons and not teams of surgeons, the model could be used for planning of minor, single surgeon, procedures such as blepharoplasties or excisions of small skin defects. Third, the within-surgeon-variability was not evaluated during this study. Random slopes for the predictive factors incorporate differences between surgeons and may have a significant influence on OR time. It is likely that the variability within a surgeon can vary and may depend on many other factors such as surgical experience, education and age.

Based on our results and the above-mentioned limitations, we have several recommendations for future research. As our primary aim was to make an inventory and quantify the effects of individual surgeons and anesthesiologists on OR time, future studies should externally validate its value for prediction models of OR time. We showed that the mean precision of a prediction model could be improved, albeit minimally. However, as this concerns the mean reduction in overtime and idle time at the center of the distribution, it is likely that larger gains were achieved in the tails of the distribution. If modeling surgeons and anesthesiologists leads to large reductions in the 90% upper prediction bounds, this could affect daily decision-making and planning (5; 24). It is important for future studies to investigate the effect of surgeons and anesthesiologists by a separate analysis of anesthesia-controlled time and surgeon-controlled time, with separate variance components for each surgeon

94

4

individually, independent if he or she is the first or second surgeon. Further gains can be achieved by incorporating random slopes for known predictive factors (i.e. surgeon’s estimate, patient and procedure characteristics) and inclusion of not evaluated case characteristics such as ASA score, anesthetic method or use of operating microscope (10; 16). At last, it might be interesting to evaluate the effect of surgical nurses, as they are part of the surgical team as well and may effect OR time (6; 29; 30).

In conclusion, this work quantified the amount of variability in OR time caused by differences between the individual surgeons and anaesthesiologists. In comparison with the type of procedure, differences between surgeons account for a small part of OR time variability. The impact of differences between anaesthesiologists on OR time is negligible. A prediction model incorporating the individual surgeons and anaesthesiologist has an increased precision but improvements are likely too marginal to have practical consequences for OR scheduling.

95

Effe

ct o

f Sur

geon

s an

d A

nest

hesi

olog

ists

on

Ope

ratin

g R

oom

Tim

e

REFERENCES

1. Eijkemans MJC, van Houdenhoven M, Nguyen T, et al. Predicting the unpredictable: a new prediction model for operating room times using individual characteristics and the surgeon’s estimate. Anesthesiology. 2010;112:41 - 49.

2. Kougias P, Tiwari V, Barshes R N, et al. Modeling anesthetic times. Predictors and implications for short-term outcomes. JournalofSurgicalResearch. 2013;180:1-7.

3. Kayis E, Wang H, Patel M, et al. Improving prediction of surgery duration using operational and temporal factors. AMIAAnnuSympProc. 2012;2012:456-462.

4. Stepaniak PS, Heij C, Mannaerts GHH, de Quelerij M, de Vries G. Modeling Procedure and Surgical Times for Current Procedural Terminology-Anesthesia-Surgeon Combinations and Evaluation in Terms of Case-Duration Prediction and Operating Room Efficiency: A Multicenter Study. Anesthesia&Analgesia. 2009;109:1232-1245.

5. Dexter F, Ledolter J. Bayesian prediction bounds and comparisons of operating room times even for procedures with few or no historic data. Anesthesiology. 2005;103:1259-1167.

6. Kayis E, Khaniyev TT, Suermondt J, Sylvester K. A robust estimation model for surgery durations with temporal, operational, and surgery team effects. Healthcaremanagementscience. 2015;18(3):222-233.

7. Steyerberg EW: Clinicalpredictionmodels:apracticalapproachtodevelopment,validation,andupdating. New York, NY, Springer, 2009.

8. Joustra P, Meester R, Ophem H. Can statisticians beat surgeons at the planning of operations? EmpirEcon. 2012;44:1697-1718.

9. Stepaniak PS, Heij C, de Vries G. Modeling and prediction of surgical procedure times. StatisticaNeerlandica. 2010;64:1-18.

10. Strum DP, Sampson AR, May JH, Vargas LG. Surgeon and type of anesthesia predict variability in surgical procedure times. Anesthesiology. 2000;92:1454-1466.

11. Silber JH, Rosenbaum PR, Ross RN, et al. Racial disparities in operative procedure time: the influence of obesity. Anesthesiology. 2013;119:43-51.

12. van Veen-Berkx E, Bitter J, Elkhuizen SG, et al. The influence of anesthesia-controlled time on operating room scheduling in Dutch university medical centres. Canadianjournalofanaesthesia=Journalcanadiend’anesthesie. 2014;61(6):524-532.

13. Silber JH, Rosenbaum PR, Zhang X, Even-Shoshan O. Influence of patient and hospital characteristics on anesthesia time in medicare patients undergoing general and orthopedic surgery. Anesthesiology. 2007;106(2):356-364.

14. Dexter F, Dexter EU, Ledolter J. Influence of procedure classification on process variability and parameter uncertainty of surgical case durations. Anesthesia and analgesia. 2010;110(4):1155-1163.

15. Cassera MA, Zheng B, Martinec DV, Dunst CM, Swanstrom LL. Surgical time independently affected by surgical team size. AmJSurg. 2009;198(2):216-222.

16. Dexter F, Dexter EU, Masursky D, Nussmeier NA. Systematic review of general thoracic surgery articles to identify predictors of operating room case durations. Anesthesia and analgesia. 2008;106(4):1232-1241.

17. Xu R, Carty MJ, Orgill DP, Lipsitz SR, Duclos A. The teaming curve: a longitudinal study of the influence of surgical team familiarity on operative time. AnnSurg. 2013;258:953-957.

96

4

18. Wright IH, Kooperberg C, Bonar BA, Bashein G. Statistical modeling to predict elective surgery time. Comparison with a computer scheduling system and surgeon-provided estimates. Anesthesiology. 1996;85:1235-1245.

19. Macario A, Dexter F. Estimating the duration of a case when the surgeon has not recently scheduled the procedure at the surgical suite. Anesthesiaandanalgesia. 1999;89(5):1241-1245.

20. Strum DP, May JH, Sampson AR, Vargas LG, Spangler WE. Estimating times of surgeries with two component procedures: comparison of the lognormal and normal models. Anesthesiology. 2003;98(1):232-240.

21. Maxwell SE, Delaney HD: Designing Experiments and Analyzing Data. PsychologyPress, 2004.

22. Duan N. Smearing Estimate: A Nonparametric Retransformation Method. JournaloftheAmericanStatisticalAssociation. 1983;78(383):605-610.

23. Bates DM: lme4: Mixed-effects modeling with R. URLhttp://lme4r-forger-projectorg/book, 2010.

24. Dexter F, Epstein RH, Traub RD, Xiao Y. Making management decisions on the day of surgery based on operating room efficiency and patient waiting times. Anesthesiology. 2004;101(6):1444-1453.

25. Dexter F, Macario A, Lubarsky DA, Burns DD. Statistical method to evaluate management strategies to decrease variability in operating room utilization: application of linear statistical modeling and Monte Carlo simulation to operating room management. Anesthesiology. 1999;91(1):262-274.

26. Ballantyne GH, Ewing D, Capella RF, et al. The learning curve measured by operating times for laparoscopic and open gastric bypass: roles of surgeon’s experience, institutional experience, body mass index and fellowship training. ObesSurg. 2005;15:172-182.

27. Stepaniak PS, Vrijland WW, de Quelerij M, de Vries G, Heij C. Working with a fixed operating room team on consecutive similar cases and the effect on case duration and turnover time. Archivesofsurgery. 2010;145(12):1165-1170.

28. Thome MA, Ehrlich D, Koesters R, et al. The point of conversion in laparoscopic colonic surgery affects the oncologic outcome in an experimental rat model. SurgEndosc. 2008;23:1988-1994.

29. Healey AN, Undre S, Vincent CA. Defining the technical skills of teamwork in surgery. QualSafHealthCare. 2006;15(4):231-234.

30. Xiao Y, Jones A, Zhang BB, et al. Team consistency and occurrences of prolonged operative time, prolonged hospital stay, and hospital readmission: a retrospective analysis. World JSurg. 2015;39(4):890-896.

31. Bates D, Maechler M, Bolker B, Walker S. Fitting linear mixed-effects models using lme4. JStatSoftw. 2015;67:1-8

97

Effe

ct o

f Sur

geon

s an

d A

nest

hesi

olog

ists

on

Ope

ratin

g R

oom

Tim

e

Interventional Studies

5Successful Interventions to Reduce First-Case

Tardiness in Dutch University Medical Centers:

Results of a Nationwide Operating Room

Benchmark Study

Elizabeth van Veen-Berkx, MScSylvia G. Elkhuizen, PhDCor J. Kalkman, MD, PhDWolfgang F. Buhre, MD, PhDGeert Kazemier, MD, PhDfor the Dutch Operating Room Benchmarking Collaborative

AmericanJournalofSurgery.2014.207(6):949-59.

102

5

ABSTRACT

Background: First-case tardiness is still a common source of frustration. In this study a nationwide operating room (OR) Benchmark database was used to assess the effectiveness of interventions implemented to reduce tardiness and calculate its economic impact.

Methods: Data from eight University Medical Centers over seven years were included: 190,295 elective inpatient first cases. Data were analyzed with SPSS Statistics and multidisciplinary focus-group study meetings. ANOVA with contrast analysis measured the influence of interventions.

Results: 7,094 hours were lost annually to first-case tardiness, which has considerable economic impact. Four UMCs implemented interventions and effectuated a significant reduction in tardiness. E.g. providing feedback directly when ORs started too late, new agreements between OR and ICU departments concerning ‘ICU bed release’ policy, and a shift in responsibilities regarding transport of patients to the OR.

Conclusions: Nationwide benchmarking can be applied to identify and measure the effectiveness of interventions to reduce first-case tardiness in a university hospital OR environment. The implemented interventions in four centers were successful in significantly reducing first-case tardiness.

103

Inte

rven

tions

to R

educ

e Fi

rst-c

ase

Tard

ines

s

INTRODUCTION

Operating rooms (ORs) are of paramount importance to a hospital, given the fact that more than 60% of patients admitted to a hospital are treated in the OR1. Efficient use of OR capacity is pivotal since it is considered a high-cost environment but a limited hospital resource2. Due to the aging population and various developments in surgery, demands for OR facilities are likely to increase2. Moreover, due to shortages of qualified OR staff, optimal utilization of ORs is an ever-increasing challenge1.

In ORs, however, inefficiencies can occur at several different moments during the day. They can occur before, during, between and after cases3,4. First-case tardiness (a ‘late start’ of the first surgical case of the day) is a common source of frustration for patients, management, and the surgical team. Once a case is delayed, a typical ‘trickle down’ effect causes the delay to increase as the day progresses, potentially affecting the rest of the scheduled patients5. This might result in cases finishing late and over-utilization of OR time. Patient satisfaction may be reduced if cases are delayed beyond their scheduled start times, particularly if patients who had to fast are kept waiting for several hours. Cases scheduled later in the afternoon may even be cancelled as a result6. This encouraged researchers to study factors that cause first-case tardiness6-8. While the majority of previous research focused on the origin of first-case tardiness, very few practical solutions to the problem have been studied5,7,9.

In 2004, the OR departments of all eight University Medical Centers (UMCs) in the Netherlands established a benchmarking collaboration, which has been active up to today. The objective of the collaboration is to improve OR performance by learning best practices from each other. Each UMC provides data on all surgical cases performed in their center to a central OR Benchmark database. Every two months multidisciplinary focus-group study meetings are organized to discuss the results of the data analysis and explore processes and practices ‘behind the data’. Through promoting dialogue between UMCs a learning environment is created. Furthermore, a national invitational conference is organized once per year to provide a broader learning and knowledge sharing platform. In comparison with the number of professionals attending the focus-group study meetings (approximately 25 to 30 professionals per meeting from all eight UMCs), these annual conferences are visited by approximately 200 professionals.

The central OR Benchmark database – today containing more than one million records of surgical cases – is used to calculate key performance indicators of the utilization of OR capacity, for example raw utilization, turnover time, under- and over-utilized OR time, and first-case tardiness. These indicators are shared between UMCs, which enables the identification of areas of improvement by comparing one’s own performance to that of other similar organizations.

This extensive database is also used for multicenter research on OR scheduling topics and OR efficiency. In the current study we aim to assess the effectiveness of interventions

104

5

Raw utilization(i.e. all case durations)

First-casetardiness

cumulativeTurnover

timeEmpty OR time Over-utilized time

Utilized OR time

Over-utilized time

Non-operative time

one OR day(in general) eight hour block time allocated to a specific surgical department

implemented to reduce first-case tardiness in a university hospital setting and to calculate the economic impact of first-case tardiness.

METHODS

All eight UMCs in the Netherlands provided data to the central OR Benchmark database on all surgical cases performed at those institutions. If an OR complex of a single UMC was divided into a main location and sub locations such as a Cancer Center, Children’s Hospital and Thorax Center, merely the main (largest) inpatient OR location was included. Longitudinal data collection within the OR benchmarking collaboration started in 2005 and is still performed today. An independent data management center administers the central OR Benchmark database. This center provides professional expertise to facilitate the collection and processing of data records. Subsequent to the collection procedure this center performs reliability checks prior to data analysis. Data provided by the data management center were used to calculate key performance indicators of the utilization of OR capacity.

The performance of one OR day, which is generally equal to eight hours of block time allocated to a specific surgical department, is commonly evaluated by the indicator ‘raw utilization’. The time when there is no patient present in the OR, so-called ‘non-operative time’, can be evaluated by three performance indicators: first-case tardiness, turnover time and empty operating room time at the end of the day, if cases finish earlier than scheduled. If cases run longer than the regularly scheduled hours of allocated block time, this is termed over-utilized time. All these performance indicators were calculated once per OR day. See Figure 1.

This study focused on first-case tardiness. Data analyzed in this research project were retrieved from the central OR Benchmark database from 1 January 2005 through 31

Figure 1. Indicators to measure the performance of one OR day

105

Inte

rven

tions

to R

educ

e Fi

rst-c

ase

Tard

ines

s

December 2011. All elective in-patient surgical cases were included. Day care surgery cases as well as non-elective (emergency) cases were excluded from the study. At the start of the collaboration, data definitions of time intervals were harmonized among all benchmarking participants10.

In 2005, first-case tardiness was already a known problem in Dutch UMCs and a common source of inefficiency for OR management, the surgical team and patients. This was the primary reason why first-case tardiness was used as a key performance indicator in the OR benchmarking. Three UMCs (UMC2, UMC3 and UMC5) decided to focus their specific efforts on the implementation of interventions to reduce first-case tardiness. UMC8 decided to focus on the implementation of a multidisciplinary preoperative team briefing in the holding area, not intentionally aiming at the reduction of tardiness, however expecting this could be a beneficial side effect since non-availability of specific team members is a known cause of tardiness3,5,8. The other four UMCs prioritized different topics for their agenda, e.g. improving OR scheduling and/or reducing turnover time.

Since 2005 (when data collection started), two focus-group study meetings concentrated solely on first-case tardiness. During these meetings it is custom to openly display the data as well as data analysis results to all benchmarking participants, to provide checks on integrity of the data and to support discussions on the interpretation of data. During these two specific meetings the interventions that were (going to be) implemented to reduce tardiness were identified and discussed. In 2012 an additional focus-group study meeting was organized around the same topic and the longitudinal data analysis was discussed in order to determine whether the interventions had proven their success over the past few years. This analysis consisted of exploring the data concerning first-case tardiness of all eight UMCs from 1 January 2005 through 31 December 2011 using descriptive statistics and box-and-whisker plots. In this particular meeting OR managers, anesthesiologists, surgeons, anesthesia nurses, OR nurses and staff advisors of all eight UMCs were represented (N = 27).

The performance indicator ‘first-case tardiness’ was defined by the difference between the scheduled starting time (generally 8:00 AM) and the actual room entry time of the first patient on that day (per operating room). This value was zero if the case entered the OR early or exactly on the scheduled time11. First-case tardiness is measured once per OR day. The common scheduled starting time was adjusted in case of an intentionally altered starting time. Every minute of first-case tardiness was calculated, as well as the percentage of first cases starting at least five minutes too late. The actual room entry time was prospectively and electronically registered by the OR nursing staff in the Hospital Information System in all individual UMCs during the operation and validated by the surgeon and anesthesiologist in charge after completion of the operation.

106

5

Statistical analysis regarding interventionsData analysis was performed with SPSS Statistics 20. Normality of distribution was determined using the Kolmogorov-Smirnov test. First-case tardiness was analyzed with the following descriptive statistics: mean ± standard deviation (SD), median, interquartile range (IQR), and box-and-whisker plots.

The Wilcoxon-Mann-Whitney test – the nonparametric alternative of the Independent-Samples T Test – was applied to determine differences between the UMCs that did implement an improvement strategy to reduce first-case tardiness and the UMCs that did not. Every UMC that implemented an intervention was also compared to the other seven UMCs to establish the impact of each intervention separately.

To measure the influence of implemented interventions to reduce tardiness, a (quasi-experimental) time-series design was applied and multiple time periods over several years before and after the intervention were evaluated12. For that reason relevant data sets were divided into four equal periods of time. The four different periods in the time-series design were compared with an analysis of variance (ANOVA). To test if the interventions led to a reduction in first-case tardiness a contrast analysis was applied: an intervention contrast, a pre-intervention contrast as well as a post-intervention contrast were tested. Prior thereto Levene’s Test was examined. Violations of the basic ANOVA assumptions were examined. The nonparametric alternative to the one-way ANOVA, the Kruskal-Wallis one-way analysis of variance, was used to confirm parametric testing.

Economic impactTo assess the economic impact of tardiness, the sum of all lost time (sum of first-case tardiness in minutes) was calculated for every UMC per year. The economic value of time wasted due to first-case tardiness was estimated according to three scenarios (A, B, and C). Scenario A was based on a more conservative approach to OR labor cost of $3.35 per regularly scheduled minute of OR time11. In this scenario, supply costs, indirect costs, anesthesiologist fees and surgeon fees were excluded. Scenario B was based on OR costs calculated in the clinical OR department of one UMC: $13.29 per regularly scheduled minute of OR time including labor costs, supply costs, indirect costs, anesthesiologist fees and surgeon fees. In both scenario A and B, the economic cost of wasted OR time was divided by the mean number of staffed ORs in that specific year, to allow for valid comparison between all UMCs. In scenario C, occurrences of late starts that lasted at least 60 minutes and maximum 120 minutes were determined (minutes of tardiness multiplied by frequency). Subsequently this amount was divided first by 60 and then by 120, to indicate how many cases with a total procedure time of 60 minutes, and a total procedure time between 60 and 120 minutes, could have been operated on in that idle time. Total procedure time was defined as ‘patient in’ to ‘patient out’ of the OR.

107

Inte

rven

tions

to R

educ

e Fi

rst-c

ase

Tard

ines

s

Relationship first-case tardiness and raw utilizationFinally, linear regression analysis was used to identify the relationship between the single predictor variable ‘first-case tardiness’ (x) and the response variable ‘raw utilization’ (y). Adjusted R-Squared (R2) values were calculated for each UMC. The performance indicator ‘raw utilization’ (%) was defined as the total amount of time patients are present in the OR, divided by the total amount of allocated OR time per day, in a given eight hour block time (e.g. 8:00h until 16:00h) x 100%. This excluded turnover time and over-utilized OR time.

RESULTS

A total of 190,295 elective in-patient surgical cases, qualifying as first cases of an OR day, were included for analysis. Mean ± SD, median, IQR and mean percentage of first cases starting at least five minutes too late per UMC during the years 2005 up to and including 2011 are shown in Table 1. These descriptive statistics demonstrated that on an overall level of eight UMCs in the Netherlands, 43% of all first operations start at least five minutes later than scheduled and that 425,612 minutes (7,094 hours or 887 eight-hour OR days) were lost to this annually. For all in-patient elective first cases of all eight UMCs, first-case tardiness showed seven minutes of reduction in IQR from 23 minutes in 2005 to 16 minutes in 2011 (Figure 2). Data of each UMC and each year showed that first-case tardiness was not normally distributed (Kolmogorov-Smirnov test, P< 0.0005) and skewness values confirmed a positively-skewed lognormal distribution.

Interventions to reduce first-case tardinessUMC2 implemented a comprehensive intervention to reduce first-case tardiness in 2007. This intervention effectuated a seven minutes reduction in IQR from 18 minutes in 2007 to 11 minutes in 2008 and the following years. Firstly, a specific team was assigned to provide feedback directly when ORs started too late, in person and on the spot every morning, by walking around. Team members consisted of an OR coordinator, anesthesiologist, surgeon, OR nurse and anesthesia nurse. Secondly, a change in activities concerning the patient process was realized: the OR nurse, instead of the anesthesia nurse, became responsible for the transport of a patient from the holding area to the OR. Meanwhile, the anesthesia nurse could continue preparing the OR for surgery. Finally, during morning hours a ‘post-call anesthesiologist’ was assigned to avoid tardiness caused by the fact that one anesthesiologist covers two ORs simultaneously. The Wilcoxon-Mann-Whitney test revealed significant differences (P< 0.0005) in first-case tardiness between UMC2 (mean rank 57,120 minutes first-case tardiness) and the other seven UMCs (mean rank 63,723 minutes).

In UMC3 the original high values of tardiness – especially in the highest 25% of the

108

5

Table 1. Descriptive statistics first-case tardiness per UMC per year

2005 2006 2007 2008 2009 2010 2011

UMC11 mean±SD tardiness, minutes 28±34 31±37 31±35 28±34 30±35 . . median tardiness, minutes 13 15 15 13 15 . . IQR (Q3-Q1) tardiness, minutes 37 38 39 37 38 . . total number of first cases, N 3,298 3,375 3,334 3,547 1,725 . . first cases starting >5 min too late, % 50% 52% 56% 54% 63% . .UMC2 mean±SD tardiness, minutes 27±47 28±48 22±35 19±35 17±30 17±30 21±35 median tardiness, minutes 9 9 8 7 7 7 9 IQR (Q3-Q1) tardiness, minutes 15 15 18 11 11 11 12 total number of first cases, N 3,472 3,501 3,402 3,156 3,203 3,380 3,412 first cases starting >5 min too late, % 38% 39% 27% 33% 39% 38% 47%UMC3 mean±SD tardiness, minutes 35±49 40±53 30±44 30±44 24±39 24±39 22±37 median tardiness, minutes 14 15 11 12 10 10 10 IQR (Q3-Q1) tardiness, minutes 35 45 27 25 20 18 15 total number of first cases, N 4,495 4,711 4,846 4,926 5,081 5,206 5,288 first cases starting >5 min too late, % 26% 25% 35% 36% 36% 39% 42%UMC4 mean±SD tardiness, minutes 33±45 32±44 33±42 45±56 48±58 38±56 40±53 median tardiness, minutes 15 15 15 17 20 13 15 IQR (Q3-Q1) tardiness, minutes 20 20 25 53 63 34 47 total number of first cases, N 2,966 3,125 2,431 3,044 3,254 3,252 3,691 first cases starting >5 min too late, % 64% 67% 70% 34% 39% 32% 42%UMC5 mean±SD tardiness, minutes 24±48 26±47 18±38 18±36 15±34 15±34 11±26 median tardiness, minutes 5 6 5 6 5 5 5 IQR (Q3-Q1) tardiness, minutes 12 15 9 10 7 7 6 total number of first cases, N 2,811 2,817 2,845 2,852 2,784 2,751 2,674 first cases starting >5 min too late, % 24% 26% 22% 26% 27% 23% 27%UMC62 mean±SD tardiness, minutes 32±46 32±45 29±40 29±42 37±56 37±57 . median tardiness, minutes 15 15 15 15 13 14 . IQR (Q3-Q1) tardiness, minutes 20 20 16 19 25 25 . total number of first cases, N 3,633 3,674 3,709 3,777 3,263 2,760 . first cases starting >5 min too late, % 47% 51% 56% 52% 43% 46% .UMC7 mean±SD tardiness, minutes 18±33 20±37 19±36 19±35 18±35 19±34 21±40 median tardiness, minutes 7 8 7 7 7 7 7 IQR (Q3-Q1) tardiness, minutes 11 11 10 10 9 10 11 total number of first cases, N 4,450 4,269 4,426 4,420 4,387 4,417 4,553 first cases starting >5 min too late, % 42% 40% 40% 40% 40% 37% 33%UMC8 mean±SD tardiness, minutes 24±36 23±35 22±32 23±29 18±28 16±27 17±29 median tardiness, minutes 10 10 11 13 9 8 7 IQR (Q3-Q1) tardiness, minutes 20 15 13 16 15 14 16 total number of first cases, N 3,313 3,275 3,379 3,179 3,358 3,293 3,423 first cases starting >5 min too late, % 57% 58% 73% 78% 49% 40% 35%

1 UMC1 data available from 2005 through June 20092 UMC6 data available from 2005 through 2010IQR = interquartile range; SD = standard deviation; UMC = university medical center

109

Inte

rven

tions

to R

educ

e Fi

rst-c

ase

Tard

ines

s

data – were caused by uncertain P ACU or ICU availability. The increased percentage of first cases starting at least five minutes too late was caused by the fact that one anesthesiologist covers two ORs simultaneously. However, tardiness was significantly reduced in 2007 (18 minutes reduction in IQR from 45 minutes in 2006 to 27 minutes in 2007 and further reduced the following years) when a new method of scheduling to control the workload of PACU and ICU departments was introduced. Furthermore, new agreements between the OR and ICU departments were implemented. Previously, early in the morning deliberation on PACU and ICU availability caused delay for the first patient scheduled for major surgery requiring post-surgical ICU. With the help of a new agreement between OR and ICU, the OR did not have to wait to start the procedure until an ICU bed was officially ‘released’. If there was no ICU capacity available, an extra temporary ICU bed was created and the OR could start without delay. Moreover, day shift starting time of anesthesia nurses was moved from 7:30 AM to 7:15 AM to generate extra time to prepare the OR. The Wilcoxon-Mann-Whitney test revealed significant differences (P< 0.0005) in first-case tardiness, however the opposite way, due to the original high value of tardiness at the starting point: UMC3 (mean rank 65,144 minutes first-case tardiness) and the other seven UMCs (mean rank 62,487 minutes).

UMC5 implemented an intervention to reduce first-case tardiness, which consisted of a shift in responsibilities: prior to 2007, the anesthesiologist had to be physically present while the anesthesia nurse brought the patient to the OR; since 2007 this was no longer

Figure 2. Reduction in first-case tardiness in minutes, all inpatient elective first cases of all 8 UMCs (n = 190,295), 2005 to 2011

2011201020092008200720062005

Fir

st-

case t

ard

iness in

min

ute

sall e

igh

t U

MC

s

130

125

120

115

110

105

100

95

90

85

80

75

70

65

60

55

50

45

40

35

30

25

20

15

10

5

0

Page 1

110

5

required due to protocol changes. Six minutes of reduction in IQR from 15 minutes in 2006 to 9 minutes in 2007 were effectuated. Since 2009 UMC5 used a new method of scheduling to control the workload of PACU and ICU departments. Furthermore, new agreements between the OR and ICU departments were implemented, similar to the ‘ICU bed release’ policy in UMC3. The Wilcoxon-Mann-Whitney test revealed significant differences (P< 0.0005) in first-case tardiness between UMC5 (mean rank 45,378 minutes first-case tardiness) and the other seven UMCs (mean rank 64,563 minutes).

In 2009 UMC8 implemented a multidisciplinary preoperative team briefing in the holding area – prior to entrance into the OR – with the objective to improve patient safety. UMC8 did not intentionally focus on the reduction of first-case tardiness; nevertheless, mean tardiness had decreased with four minutes and further reduced the following years, since the implementation of the preoperative team briefing. The Wilcoxon-Mann-Whitney test revealed no significant differences (P< 0.883) in first-case tardiness between UMC8 (mean rank 62,899 minutes first-case tardiness) and the other seven UMCs (mean rank 62,857 minutes).

Effectiveness of interventionsFigure 3 illustrates that the group of four UMCs with an intervention effectuated seven minutes of reduction in IQR from 20 minutes in 2005 to 13 minutes in 2011 (P< 0.0005, Wilcoxon-Mann-Whitney). The other group of four UMCs without an intervention showed an IQR of approximately 25 minutes of first-case tardiness every year.

Results of UMC2 and UMC8 showed a significant difference with regard to the intervention contrast (P< 0.0005). These UMCs also showed that differences in tardiness concerning the pre-intervention contrast (P < 0.566 and P < 0.105 respectively) and post-intervention contrast (P < 0.344 and P < 0.498 respectively) were not significant. UMC5 revealed significant results for the intervention contrast (P < 0.0005) as well as the post-intervention contrast (P < 0.0005); the pre-intervention contrast was not significant (P < 0.387). UMC3 demonstrated significant results for all three contrasts with P-values of respectively P < 0.0005, P < 0.007 and P < 0.0005. See Figure 4 for results of the ANOVA with contrast analysis. Levene’s Test was significant (P< 0.0005), thus equal variances between the four equal periods (two before and two after the intervention) were not assumed.

Economic impactThe total loss due to first-case tardiness calculated from the findings of this study, using scenario A and B, are shown in Table 2 per UMC and in any investigated year. All calculations were made in US dollars. Regarding economic impact, the intervention implemented in UMC2 effectuated the largest reduction of all interventions, followed by UMC8, UMC5 and UMC3, in that order. UMC2 realized a reduction of 27,392 minutes of tardiness in one year. With scenario A this meant a shrinkage of $91,763 ($6,555 per OR) and with scenario B a shrinkage of $364,040 ($26,003 per OR).

111

Inte

rven

tions

to R

educ

e Fi

rst-c

ase

Tard

ines

s

0

5

10

15

20

25

30

2005 2006 2007 2008 2009 2010 2011

IQR

fir

st-

case t

ard

iness in

min

ute

s

Four UMCs with intervention

Four UMCs without intervention

Figure 3. Significant differences in first-case tardiness (IQR in minutes) between UMCs with and UMCs without an intervention

Figure 4. Results of ANOVA with contrast analysis, four UMCs with intervention to reduce first-case tardiness

112

5

Considering the total dataset of all eight UMCs, scenario C showed 6.5% (N = 6,790) of first-case tardiness that lasted at least 60 minutes and maximum 120 minutes, which added up to a total amount of 582,437 minutes of lost time. A number of that 9,707 cases with a total procedure time of 60 minutes (a mean of 173 cases per year per UMC) or 4,854 cases with a total procedure time between 60 and 120 minutes (a mean of 87 cases per year per UMC) could have been operated on in that idle time.

Relationship first-case tardiness and raw utilizationLinear regression analysis demonstrated a significant relationship between first-case tardiness and the variation in raw OR utilization in each UMC (P< 0.0005). On an overall level of all eight UMCs, 28% of the variation in raw utilization was explained by the variation in first-case tardiness. Adjusted R-Squared (R2) values per UMC ranged from 18% to 34%. Figure 5 depicts a scatter plot of one random UMC (5) showing first-case tardiness in minutes against raw utilization%, N = 19,534 in-patient elective first cases, adjusted R2-value is 22%.

Figure 5. Scatter plot of first-case tardiness (minutes) against raw utilization (%). Data of UMC5, n = 19,534 inpatient elective first cases. Adjusted R2 value of UMC5 = 22%

113

Inte

rven

tions

to R

educ

e Fi

rst-c

ase

Tard

ines

s

Table 2. Economic impact of first-case tardiness per UMC per year, all calculations in US dollars $

2005 2006 2007 2008 2009 2010 2011

UMC1 sum of FCT in minutes 63,258 70,465 73,216 72,733 40,837 . . estimation A based on $3,35 p/m 211,914 236,058 245,274 243,656 136,804 . . $ / OR 16,301 18,158 18,867 17,404 19,543 . . estimation B based on $13,29 p/m 840,699 936,480 973,041 966,622 542,724 . . $ / OR 64,669 72,037 74,849 69,044 77,532 . .

UMC2 sum of FCT in minutes 58,365 62,058 34,666 35,530 36,401 37,833 49,791 estimation A based on $3,35 p/m 195,523 207,894 116,131 119,026 121,943 126,741 166,800 $ / OR 13,966 14,850 8,295 9,156 9,380 9,053 11,914 estimation B based on $13,29 p/m 775,671 824,751 460,711 472,194 483,769 502,801 661,722 $ / OR 55,405 58,911 32,908 36,323 37,213 35,914 47,266


UMC4 sum of FCT in minutes 77,922 82,062 67,828 59,318 77,250 54,267 82,116 estimation A based on $3,35 p/m 261,039 274,908 227,224 198,715 258,788 181,794 275,089 $ / OR 21,753 21,147 22,722 16,560 19,907 13,984 18,339 estimation B based on $13,29 p/m 1,035,583 1,090,604 901,434 788,336 1,026,653 721,208 1,091,322 $ / OR 86,299 83,893 90,143 65,695 78,973 55,478 72,755


UMC6 sum of FCT in minutes 65,690 74,122 75,858 73,236 70,592 65,271 . estimation A based on $3,35 p/m 220,062 248,309 254,124 245,341 236,483 218,658 . $ / OR 14,671 16,554 16,942 16,356 18,191 19,878 . estimation B based on $13,29 p/m 873,020 985,081 1,008,153 973,306 938,168 867,452 . $ / OR 58,201 65,672 67,210 64,887 72,167 78,859 .



114

5

DISCUSSION

On an overall level of eight UMCs in the Netherlands, 43% of all first operations start at least 5 minutes later than scheduled and 425,612 minutes are lost due to this annually, which has a respectable economic impact. This study also shows that on an overall level of all UMCs first-case tardiness has decreased since 2005; this decrease persists up to and including 2011 (study period). Moreover, this study demonstrates that four UMCs implemented successful interventions to reduce tardiness. These UMCs showed a stepwise reduction in variation of first-case tardiness, in other words a decrease in IQR during the years, which indicates an organizational learning effect17. ANOVA with contrast analysis shows that a marked change occurred at the time of the intervention in these four UMCs, which indicates the success of their interventions.

The ANOVA with contrast results of specifically UMC3 demonstrated that the trend toward improvement may have been present prior to the intervention. This finding suggests that the original high value of tardiness of UMC3 at the starting point might be an important determinant for improvement. A high sense of urgency is a critical success factor for a change process to succeed18. Purely based on the original values of tardiness from UMC2, UMC5 and UMC8, these centers had less sense of urgency and fewer room to improve first-case tardiness; nevertheless, they did and also succeeded significantly. UMC3 showed the highest relative improvement because of their lower original value and thus having more room to improve than other UMCs.

Inefficient use of OR capacity is a worldwide problem. Previous studies have been carried out with the goal to increase efficiency, allowing additional cases to be performed in the same operating time for the same cost. A number of studies focused explicitly on first-case tardiness3,5,8,9,11,19-24. However, these studies were performed within the context of either one or two (university) hospital(s), only one or two surgical services within a hospital or with the use of data collected in one year. This is the first nationwide longitudinal multicenter study that involved repeated and continuous measurement of the same parameters – including first-case tardiness – for a period of seven years and is, with 190,295 (first) surgical cases, the largest set of OR data published from the Netherlands to date.

Although recent studies have indicated that first-case tardiness does not affect OR efficiency23, 26, 27 and the ‘trickle down’ effect has been argued against23, 22, 19, 28 first-case tardiness remains of interest because it continues to be perceived as a key performance indicator of inefficiency in the OR3. Moreover, this can be confirmed in the OR practice of all eight UMCs in the Netherlands, the participants of this benchmark study as 28% of the variation in raw utilization was explained by the variation in first-case tardiness in the current study. Also other fundamental elements might be influenced by it in a negative manner. Patient satisfaction may be reduced if operations are delayed beyond their scheduled start times,

115

Inte

rven

tions

to R

educ

e Fi

rst-c

ase

Tard

ines

s

particularly if patients who had to fast are kept waiting for several hours6, 3, 7, 5. Furthermore, delays are a source of frustration for health care professionals and, although, time saved by reducing first-case tardiness cannot be accommodated with extra operations, the time saved is still time that can be used for other purposes3,8. The multidisciplinary focus group in this study corroborated that starting on time means less rush at the beginning and potentially throughout the day; and rushing has been identified as one of the factors that lead to an unsafe working environment5, 29. In this context, the outcomes of this study may contribute to the improvement of overall operating room practice.

With reference to the central OR Benchmark database and specifically the two performance indicators analyzed in this study, first-case tardiness and raw utilization, a major concern of readers could be the distribution of the data and the manner of statistical testing. Data in the recent study showed a positively-skewed lognormal distribution, thus, the assumption of normality was dishonored. However, ANOVA is considered a robust test against the normality assumption, particularly with large sample sizes (N ≥ 1,000), which was the case in this study. This is particularly true for larger sample sizes, since the sampling distributions then have weaker dependence on the shape of the population distribution13, 14, 15, 16. In addition, Kruskal-Wallis one-way analysis of variance showed the same results and therefore one-way ANOVA with contrasts was further applied in this study to compare more than two groups. Concerning linear regression analysis, normality of data is not a principal assumption. Normality of the error distribution is a principal assumption, which justifies the use of linear regression, yet again, it is not imperative for large sample sizes (N ≥ 1,000), which was the case in this study13,

14, 15, 16. Benchmarking is more than performance comparison between organizations. Our

nationwide OR benchmarking collaboration focuses mainly on learning from each other, knowledge sharing, discussing strengths and weaknesses, and identifying good practices. Multidisciplinary focus-group study meetings are frequently organized within the collaboration to discuss data analysis results and explore processes and practices ‘behind the data’. Through promoting dialogue between UMCs a learning environment is created. The focus-group study meeting within this specific research appeared to be an effective method to identify interventions that were implemented to reduce first-case tardiness (specific goal) in UMC2, UMC3 and UMC5, and also to identify a strategy to improve patient safety, while demonstrating that a reduction in first-case tardiness appeared to be an attractive side effect (in UMC8).

The current study has several limitations. First, within a health care system, particularly in a complex and dynamic environment such as the OR, multiple changes occur during any given period. The evaluation of quality improvements, like the interventions to reduce first-case tardiness in this study, frequently rely on weak “before-after” designs12. These “other” changes might have produced the preferred improvements, instead of the specific intervention. One

116

5

way to minimize this possibility, is to consider multiple time periods in a time-series design as used in this research. The applied ANOVA with three contrasts conveys the extent of background variation and also indicates the extent to which any trend toward improvement may have been present prior to the intervention12.

Second, the calculation of the specific economic value due to the loss of OR time in absolute terms also remains complex9, 20. This is particularly the case for UMCs that typically have three core responsibilities: teaching and training, research and tertiary patient care. The yearly loss in labor costs11 was estimated (scenario A), which is a rather conservative calculation method since supply costs, indirect costs, anesthesiologist fees and surgeon fees are excluded. However, critics will object to this way of calculating the economic value of losing OR time, because the extra minutes gained would not allow any additional cases to be performed5. Dexter et al.30 found that, due to a lack of knowledge and a psychological bias on this topic, OR managers can become fixated on strategies to avoid first-case tardiness. Dexter et al. state that first-case delays are small delays in time, which are not clearly economically important, because the costs of reducing these delays are often high and time reduction in each OR is often limited. In addition, Macario26, 27 suggested – however lacking data to support the claim – that first-case tardiness of up to 45 minutes remain consistent with efficient performance23. That is why, in this study, specific time intervals of tardiness were investigated. Scenario (C)estimatedtheeconomicimpactoffirst-casetardinessbyfocusingonaspecifictimeintervaloftardinessbetween60and120minutes;andfoundthat9,707 cases with a total procedure time of 60 minutes or 4,854 cases with a total procedure time between 60 and 120 minutes could have been operated on inthatidletime.This could contribute to the reduction of inefficiency.

Reducing first-case tardiness and increasing the proportion of on-time starts is merely one aspect of efficient use of OR capacity. In ORs, inefficiencies can occur before, during, between and after cases31. Further research is required considering the additional performance indicators in this nationwide multicenter Benchmark database such as turnover time, under-utilized OR time, over-utilized OR time and the difference between the estimated and actual duration of operations.

In conclusion, first-case tardiness occurs on a daily basis in Dutch UMCs and this has a sizeable impact on OR efficiency. Yet, this study shows that benchmarking can help to overcome this by exchanging best practices and prevent ‘reinventing the wheel’ through organized learning and networking. In accordance with De Korne etal.32ourstudycorroboratesthatbenchmarkingishighlydependentonsocialprocessesandalearningenvironmentparalleltoastructuredandrationalprocessofsharingperformancedata.Transfer of knowledge is one of the major targets of the OR Benchmarking collaboration. During the two-monthly organized multidisciplinary focus-group study meetings and the yearly national invitational conference, targets between hospitals are a matter of discussion and presentation. The overall data

117

Inte

rven

tions

to R

educ

e Fi

rst-c

ase

Tard

ines

s

presentation is accomplished by best practices from different hospitals. Thus, knowledge transfer is performed according to two routes: data analysis and best practice sharing.

Overall, this study shows that benchmarking can be applied to identify and measure the effectiveness of interventions to reduce first-case tardiness in a university hospital OR environment.

118

5

REFERENCES




4. Wang J, Dexter F, Yang K. A Behavioral Study of Daily Mean Turnover Times and First Case of the Day Start Tardiness. Anesth Analg 2013.


6. Wachtel RE, Dexter F. Influence of the operating room schedule on tardiness from scheduled start times. Anesth Analg 2009;108:1889-901.


8. Panni MK, Shah SJ, Chavarro C, Rawl M, Wojnarwsky PK, Panni JK. Improving operating room first start efficiency - value of both checklist and a pre-operative facilitator. Acta Anaesthesiol Scand 2013.

9. Lapierre SD, Batson C, McCaskey S. Improving on-time performance in health care organizations: a case study. Health Care Manag Sci 1999;2:27-34.

10. Van Houdenhoven M, Hans EW, Klein J, Wullink G, Kazemier G. A norm utilisation for scarce hospital resources: evidence from operating rooms in a Dutch university hospital. J Med Syst 2007;31:231-6.


12. Shojania KG, Grimshaw JM. Evidence-based quality improvement: the state of the science. Health Aff (Millwood) 2005;24:138-50.

13. Larson MG. Analysis of variance. Circulation 2008;117:115-21.

14. Neill J. Writing Up An ANOVA Analysis. Center for Applied Psychology, University of Canberra 2007.




18. Kotter JP. Leading change. Boston, Mass.: Harvard Business Review Press; 2012.

19. Dexter EU, Dexter F, Masursky D, Garver MP, Nussmeier NA. Both bias and lack of knowledge influence organizational focus on first case of the day starts. Anesth Analg 2009;108:1257-61.

20. Ernst C, Szczesny A, Soderstrom N, Siegmund F, Schleppers A. Success of commonly used operating room management tools in reducing tardiness of first case of the day starts: evidence from German hospitals. Anesth Analg 2012;115:671-7.

21. Fezza M, Palermo GB. Simple solutions for reducing first-procedure delays. AORN J 2011;93:450-4.

119

Inte

rven

tions

to R

educ

e Fi

rst-c

ase

Tard

ines

s

22. McIntosh C, Dexter F, Epstein RH. The impact of service-specific staffing, case scheduling, turnovers, and first-case starts on anesthesia group and operating room productivity: a tutorial using data from an Australian hospital. Anesth Analg 2006;103:1499-516.


24. Windle PE, Barron K, Walker D, Cormier J. A COMIT model utilization to improve first-case start time. Lippincotts Case Manag 2001;6:38-46.

26. Macario A. Are your hospital operating rooms “efficient”? A scoring system with eight performance indicators. Anesthesiology 2006;105:237-40.

27. Macario A. The limitations of using operating room utilisation to allocate surgeons more or less surgical block time in the USA. Anaesthesia 2010;65:548-52.

28. Dexter F, Macario A. When to release allocated operating room time to increase operating room efficiency. Anesth Analg 2004;98:758-62.

29. Greenberg CC, Roth EM, Sheridan TB, et al. Making the operating room of the future safer. Am Surg 2006;72:1102-8; discussion 26-48.

30. Dexter F, Lee JD, Dow AJ, Lubarsky DA. A psychological basis for anesthesiologists’ operating room managerial decision-making on the day of surgery. Anesth Analg 2007;105:430-4.


32. de Korne DF, Sol KJ, van Wijngaarden JD, et al. Evaluation of an international benchmarking initiative in nine eye hospitals. Health Care Manage Rev 2010;35:23-35.

6Scheduling Anesthesia Time Reduces Case

Cancellations and Improves Operating Room

Workflow in a University Hospital Setting

Elizabeth van Veen-Berkx, MSc Menno V. van Dijk, MScDiederich C. Cornelisse, MScGeert Kazemier, MD, PhDFleur C. Mokken, MD, PhD

JournalAmericanCollegeofSurgeons.2016.223:343-51.

122

6

ABSTRACT

Background: A new method of scheduling anesthesia-controlled time (ACT) was implemented on 1 July 2012 in an academic inpatient OR department. This study examines the relationship between this new scheduling method and operating room (OR) performance. The new method comprised of the development of predetermined time frames per anesthetic technique based on historical data of the actual time needed for anesthesia induction and emergence. Seven so-called ‘anesthesia scheduling packages’ (0 – 6) were established. Several options based on the quantity of anesthesia monitoring and the complexity of the patient were differentiated in time within each package.

Study Design: Quasi-experimental time-series design. Relevant data divided into four equal periods of time. These time periods were compared with an ANOVA with contrast analysis: an intervention, preintervention and postintervention contrast were tested. All emergency cases were excluded. A total of 34,976 inpatient elective cases performed during the time period of 1 January 2010 to 31 December 2014 were included for statistical analyses.

Results: The intervention contrast showed a significant decrease (P < 0.001) of 4.5% in the prediction error. The total number of cancellations reduced with 19.9%. The ANOVA with contrast analyses showed no significant differences with respect to under- and overutilized OR time and raw utilization. Unanticipated results derived from this study, allowing for a smoother workflow: e.g. anesthesia nurses know exactly which medical equipment and devices need to be assembled and tested beforehand, based on the scheduled anesthesia package.

Conclusions: Scheduling the two major components of a procedure (anesthesia- as well as surgeon-controlled time) more accurately, leads to less case cancellations, lower prediction errors and smoother OR workflow in a university hospital setting.

123

Ane

sthe

sia

Tim

e an

d O

pera

ting

Roo

m S

ched

ulin

g

INTRODUCTION

Hospital care is still faced with the challenge to provide safe, high quality care, and at the same time the need to control costs. Operating room (OR) departments are amidst the most important departments of a hospital, accounting for a considerable amount of workload, revenues, as well as costs. For this reason, OR inefficiencies should be avoided where possible. Besides, OR inefficiencies are a dissatisfier for all – and there are many – parties involved. One way to improve OR efficiency is to optimize surgical case scheduling.

Several previous studies1-11 have concentrated on the prediction of total procedure time. Total procedure time is subdivided into anesthesia induction time, surgeon-controlled time (SCT, including patient positioning, prepping and draping) and anesthesia emergence time (Figure 1). The sum of induction time and emergence time is also known as anesthesia-controlled time (ACT). In the Netherlands, the overall current prediction method is as follows: the surgeon’s prediction of SCT is determined before each procedure. In some hospitals, surgeons make a routine prediction of the time needed, and in others, historical times are the point of reference5, 11, 12. Yet, the exactness of these predictions is limited13. For the prediction of ACT, usually a fixed time period of e.g. 20 minutes (for general anesthesia) or 40 minutes (for a regional anesthetic technique) is added to the surgeon’s prediction of SCT. Together this provides the predicted total procedure time used for OR scheduling14.

Previous studies predominantly focused on the subject of predicting the time frame employed by surgeons, which accounts for the major part of total procedure time. However, in a former study, we found that in a university hospital setting, a minimum of 25% up to 30% of total procedure time is engaged by anesthesiologists14. In the Netherlands, university

Startcase

Total Procedure Time (one session/one case)


Surgeon-Controlled Time (SCT)Anesthesia

Emergence Time

End case/start next case

Anesthesia-Controlled Time (ACT)


Surgeon-Controlled TimeAnesthesia

Emergence Time

Figure 1. Total procedure time is subdivided into anesthesia induction time, surgeon-controlled time, and anesthesia emergence time. The sum of induction and emergence time is anesthesia-controlled time.

124

6

medical centers (UMCs) take responsibility for tertiary care, clinical research, innovation, and training & teaching. The complexity of patients with multiple comorbidities, as well as the complexity of surgical procedures, usually results in a longer duration of surgery than in general hospitals. This was also validated in a Dutch, nationwide OR benchmark study in which eight UMCs participate, based on 330,258 inpatient elective surgical cases14: the mean (SD) Total Procedure Time of 158 (119) minutes and the median of 124 minutes reflect that the complexity of procedures is potentially greater than in other (general) hospitals1. This level of complexity of the patient case mix in UMCs can make it more difficult to accurately predict their duration and complicate efficient scheduling. Based on this OR benchmark database, further results affirmed that ACT is a considerable component of total procedure time, which should be scheduled just as realistically as SCT. Therefore, we advised that grossing up the SCT by 33% to account for ACT, rather than scheduling a fixed number of minutes, improves the prediction of total procedure time14.

Even though we demonstrated in this former study that this recommended scheduling rule leads to more prediction accuracy, nevertheless, a ‘scheduling deviation’ remains. One Dutch UMC, the Academic Medical Centre (AMC) in Amsterdam, adopted a new system of scheduling ACT based on predetermined time frames per anesthetic technique. Previous studies suggest that more accurate prediction rules may lead to reducing the amount of under- and overutilized OR time, as well as the number of case cancellations14-17. For that reason, this recent study aims to examine the relationship between this new scheduling method and OR performance.

New OR scheduling method anesthesia-controlled time (the intervention)The AMC is a university hospital affiliated with the University of Amsterdam. It has an intensive cooperation with the other university hospital of Amsterdam, the VU University Medical Centre (VUmc). Like the other seven UMCs in the Netherlands, the AMC offers a “last resort” (tertiary care) function for patients with complex healthcare issues and combines this top-level patient care with research, training and education.

In AMC Amsterdam the OR management team decided to implement a new strategy with regard to realistic scheduling. This new strategy comprised of the development of predetermined time frames per anesthetic technique based on historical data of the actual time needed for anesthesia induction and emergence. In total seven so-called ‘anesthesia scheduling packages’ (0 – 6) were established (Table 1). Several options based on the quantity of anesthesia monitoring (e.g. intubation, arterial line, central line) and the complexity of the patient were differentiated in time within each package. During the pre-anesthesia check-up the anesthesiologist assigns and enters the package required for each specific patient

1 In the Netherlands, the mean Total Procedure Time of surgical cases performed in general hospitals, is 74 minutes. This number is based on a Dutch benchmark, specifically for general hospitals, organized and hosted by a consultancy firm. In total twelve general hospitals participate in this benchmark.

125

Ane

sthe

sia

Tim

e an

d O

pera

ting

Roo

m S

ched

ulin

g

and procedure in the Hospital Information System. Surgeons use the same system for OR scheduling and are thus aware of the time frame already scheduled for ACT. Additionally, the surgeon schedules the SCT needed, including positioning, skin preparation and draping.

Afterwards, the actual ACT is registered per anesthesia package and per anesthesiologist, which is similar to the registration of actual SCT (per procedure and per surgeon). This way, individual historical data are obtained and time frames used for scheduling can evolve.

Table 1. AMC Anesthesia scheduling packages with predetermined time frame (including induction and emergence)

126

6

METHODS

The new method of scheduling ACT based on predetermined time frames per anesthetic technique, appointed as ‘the intervention’, was implemented on 1 July 2012. To measure the influence of this intervention, a (quasi-experimental) time-series design was applied and multiple time periods over two and a half years before and two and a half years after the intervention were evaluated18.

Study populationTo define a consistent dataset for analysis, we included all medical departments performing surgery in the inpatient OR department of the AMC. All emergency cases were excluded since emergency cases are not subjected to the preliminary scheduling process. All surgical cases with a total procedure time and/or a SCT of less than 5 minutes or more than 1,440 minutes were excluded to eliminate potential flaws registration errors. A total of 34,976 inpatient elective cases performed during the time period of 1 January 2010 to 31 December 2014 were included for statistical analyses.

Organizational characteristics: total procedure timeAccording to Dexter19, anesthesia-controlled time is defined as ‘‘the sum of (1) the time starting when the patient enters the OR to the time when surgical positioning or skin preparation can begin plus (2) the time starting when the surgical dressing is completed and ending when the patient leaves the OR,’’ in other words, ACT is the sum of (1) anesthesia induction time plus (2) anesthesia emergence time. Dexter also defined SCT19 as ‘‘the time starting when patient positioning and/or skin preparation can begin to when surgical dressing is completed’’ (Figure 1). Total procedure time or case duration (in minutes) was defined as the time from the patient’s entry into the OR until the patient’s departure from the OR, i.e. ACT plus SCT. It is also referred to as ‘one OR session’.

OR Performance IndicatorsSeveral indicators to evaluate OR performance were considered in this study: empty OR time at the end of the day (minutes), over-utilized time or overtime (minutes), the number of case cancellations (absolute numbers), raw utilization (%), and prediction error (%).

Empty OR time at the end of the day was quantified by the difference in minutes between the actual room exit time of the last patient and the scheduled end of block time (4:30 PM), finishing before 4:30 PM17, 20. The common end of block time was adjusted in case of an intentionally extended allocated block time, which is more than the standard of eight and a half hours (e.g. 6:00 PM).

127

Ane

sthe

sia

Tim

e an

d O

pera

ting

Roo

m S

ched

ulin

g

Table 2. The absolute number of cancellations per time period, for all different reasons/causes that were recorded in AMC

Reason/cause for cancellation

Time period and number of cancellations

I II III IV

N N N N

due to a lack of available time on the OR schedule at the end of the day

418 437 367 320

emergency case during elective list 168 144 143 108

OR team not complete 92 60 48 61

reason was not obtained/registered 71 71 77 68

medical condition change 55 59 62 37

lack of evident surgical indication 50 60 55 54

patient developed medical illness 44 34 39 36

due to calamities 24 2 0 1

surgeon not available 23 32 12 12

patient not fasted 23 24 21 23

administrative reasons 16 21 31 24

patient moved from outpatient to inpatient OR 16 16 6 14

patient refuses surgery 12 22 9 18

inadequate preoperative condition 11 13 17 7

no-show 11 21 5 8

awaiting additional diagnostics 9 11 15 27

patient already operated (due to emergency reasons) 9 13 15 10

ICU bed not available 7 2 4 1

lack of relevant OR equipment 7 5 8 10

incorrect registration 5 6 6 11

alteration in elective list during the week 5 5 0 0

cancelled at patient’s request 3 4 0 1

lack of necessary (sterile) instruments 2 1 0 0

ward bed not available 2 1 0 1

patient moved from inpatient to outpatient OR 2 1 1 2

PACU bed not available 1 2 0 2

no-show at pre-op assessment 1 0 0 2

patient not insured 1 1 1 0

recovery bed not available 0 2 0 0

patient moved to another hospital 0 2 5 3

patient deceased 0 1 1 0

patient cannot be reached (no-show) 0 3 1 1

stopping rule in preoperative process/checklist 0 0 0 1

patient back on waiting list at surgeon’s request 0 1 0 0

Total number of cancellations 1.088 1.077 949 863

128

6

Time period

IVIIIIII

Mean

pre

dic

tio

n e

rro

r (%

) to

tal p

roced

ure

tim

e

17

16

15

14

13

12

Page 1

Figure 2. Significant reduction in mean prediction error (%), significant intervention contrast (P < .001).

Overtime at the end of the day was quantified by the difference in minutes between the scheduled end of block time (4:30 PM) and the actual room exit time of the last patient, finishing after 4:30 PM21. The common end of block time was adjusted in case of an intentionally extended allocated block time.

A cancellation on the day of intended surgery was defined as an operation that was scheduled on the final, elective OR schedule for that day but was not performed on that day. In other words, the date of cancellation was equal to the date of intended surgery22-24. Each cancellation with an associated reason was registered in the Hospital Information System of the AMC. It is common practice in ORs to register the reason for every cancellation, however, these reasons are not standardized and differ per hospital. In Table 2 and Figure 3 (in the Results section) all reasons that were recorded in AMC are explicated. These reasons can be divided into categories, such as: administrative, medical, patient-related, hospital-related. Cancellations associated with hospital-related reasons were hospital-initiated cancellations that were due to inefficiencies in the organizational system and that were potentially avoidable23.

The absolute number of cancellations was evaluated on different levels: on a total level, in detail for every recorded reason, as well as specifically for the reason “due to a lack of available time on the OR schedule at the end of the day (due to overtime of the previous case)”. This specific reason is potentially avoidable.

The overall performance of one OR day, which is generally equal to 8,5 hours of block time allocated to a specific surgical department, is universally expressed as the indicator “raw utilization.” Raw utilization was defined as the total amount of time patients are present in the

129

Ane

sthe

sia

Tim

e an

d O

pera

ting

Roo

m S

ched

ulin

g

OR, divided by the total amount of allocated block time per day ˟ 100%. This definition of raw utilization excluded turnover time and overused OR time25.

The prediction error (%)14, 21 was defined as the difference between the actual and the expected total procedure time (used for OR scheduling), divided by the expected total procedure time ˟ 100%.

Statistical analysisData analysis was performed using SPSS statistics software (version 21, IBM SPSS). Normality of distribution was determined using the Kolmogorov-Smirnov test. Total procedure time and the performance indicators prediction error (%), raw utilization (%), underused OR time (minutes) and overtime (minutes) were analysed with the following descriptive statistics: mean (SD), median, interquartile range, and box-and-whisker plots. To measure the influence of the intervention, which was implemented on 1 July 2012, a quasi-experimental time-series design was applied and therefore relevant data was divided into four equal periods of time of 15 months18, with two measurement periods before the implementation and two periods after the implementation:

Two measurement periods before the implementation:

Two measurement periods after the implementation:

01-01-2010 to 31-03-2011 (I); 01-07-2012 to 30-09-2013 (III);01-04-2011 to 30-06-2012 (II). 01-10-2013 to 31-12-2014 (IV).

Those four time periods were compared with an analysis of variance (ANOVA) with contrasts. To test whether changes in performance indicators were initiated by the intervention, three contrasts were considered:

a) an intervention contrast (comparing time periods I and II before the intervention with time periods III and IV after the intervention),

b) a before-intervention contrast (comparing time periods I and II before the intervention), and

c) an after-intervention contrast (comparing time periods III and IV after the intervention).

To attribute a difference in OR performance to the intervention per se, the expectation was that the intervention contrast was significant (P < 0.01) and both the before- and after-measurement contrasts were not significant (P > 0.01).

Prior thereto, Levene’s test was examined. Violations of the basic ANOVA assumptions were examined. The nonparametric alternative to the one-way ANOVA, the Kruskal-Wallis one-way ANOVA was used to confirm parametric testing. Finally, the progress of the absolute number of cancellations was analysed with regard to the same four time periods.

130

6

RESULTS

Organizational characteristics: total procedure timeA total of 34,976 inpatient elective cases, performed during the time period of 1 January 2010 to 31 December 2014, were selected for inclusion in this study. Descriptive statistics demonstrated that, on an overall level for all surgical specialties and the total amount of inpatient elective cases, mean total procedure time at the beginning of the study period was 161 minutes (SD 113). At the end of the study period mean total procedure time was 186 minutes (SD 127). When the increase of total procedure time was evaluated over time with an ANOVA contrast analysis, interestingly, a significant difference showed after the new OR scheduling method was implemented on 1 July 2012. In this specific situation, the before-measurement contrast was not significant but the intervention contrast as well as the after-measurement contrast were significant. This indicates that total procedure time, more specifically surgeon-controlled time, was continuously increasing during the complete study period (Table 3).

Table 3. Descriptive statistics of total procedure time, surgeon-controlled time and anesthesia-controlled time per time period. Time periods I and II: before-intervention (data of 2,5 years); time periods III and IV: after-intervention (data of 2,5 years).

Total procedure time

Surgeon-controlled time

Anesthesia-controlled time

Time Period N Mean SD Mean SD Mean SD

I 8,577 161 113 108 96 35 22

II 8,544 165 117 110 99 37 28

III 8,888 174 124 118 106 36 27

IV 8,962 186 127 128 109 37 27

Total 34,971 172 121 116 103 36 26

OR Performance IndicatorsTo attribute a difference in OR performance to the intervention per se, the expectation was that the intervention contrast was significant (P < 0.01) and both the before- and after-measurement contrasts were not significant (P > 0.01). In all three contrasts, there were no significant differences with respect to raw utilization, empty OR time and overtime at the end of the day, detected in the ANOVA with contrast analyses.

Figure 2 demonstrates the decrease in the prediction error (%) from almost 17% prior to the intervention to 12.5% after the intervention. The before- and after-intervention contrasts

131

Ane

sthe

sia

Tim

e an

d O

pera

ting

Roo

m S

ched

ulin

g

Figu

re 3

. The

abs

olut

e nu

mbe

r of c

ance

llatio

ns p

er ti

me

perio

d, fo

r all

diffe

rent

reco

rded

reas

ons/

caus

es

Tim

e pe

riods

I an

d II:

bef

ore

the

inte

rven

tion

(dat

a of

2,5

yea

rs)

Tim

e pe

riods

III a

nd IV

: afte

r the

inte

rven

tion

(dat

a of

2,5

yea

rs)

132

6

showed no significant differences, however a significant decrease of 4.5% was confirmed for the intervention contrast (P < 0.001), indicating that the decrease can be attributed to the new scheduling method.

The total number of cancellations reduced from 1,077 before the intervention to 863 after the intervention, which is a decrease of 19.9%. Table 2 and Figure 3 show the absolute number of cancellations for all different recorded reasons. From the data in this figure, it is apparent that the main reason for cancellation was “due to a lack of available time on the OR schedule at the end of the day (due to overtime of the previous case)”, followed by “emergency case during elective list”. The number of cancellations due to a lack of OR time reduced from 437 before-intervention to 367 (-16%) after-intervention, and later on reduced even further to 320; a sizeable reduction of 26.8% compared to the time period before the intervention.

DISCUSSION

This study shows that the introduction of a novel scheduling method for anesthesia-controlled time results in reduced prediction errors and fewer case cancellations. Simultaneously, the mean total procedure time increased. An interesting finding is, that the number of cancellations specifically “due to a lack of available time on the OR schedule at the end of the day” declined, which is the single cause that might be attributable to the change in scheduling methodology.

These findings provide important implications with respect to OR scheduling in a university hospital setting, since they affirm that anesthesia time is a considerable component of total procedure time and should be scheduled just as realistically as and separate from surgeon-controlled time. Scheduling the two major components of a procedure (ACT as well as SCT) more accurately, results in less case cancellations and lower prediction errors. This may lead to more patient satisfaction and a more efficient use of limited and expensive OR resources.

The recent research builds on and supports a former Dutch multicentre study that investigated a more theoretical approach of OR scheduling, based on a comprehensive OR benchmark dataset14. This multicentre study already claimed that respecting the variability in ACT with more accurate and realistic prediction rules is preferred over employing the general methodology based on a fixed number of minutes (e.g. 20 minutes). The current findings match those observed by Escobar etal.26 who found significant variation in anesthesia release time and concluded that assigning a constant fixed time for anesthesia induction is inappropriate for OR scheduling purposes.

Two former prospective, observational studies showed that surgeons and anesthesiologists are not always capable to predict the necessary time for a procedure4, 27. Anesthesiologists performed even lower than surgeons and underestimated the time for induction with 35 minutes. It was also clearly demonstrated that induction times in elderly, high-risk patients who

133

Ane

sthe

sia

Tim

e an

d O

pera

ting

Roo

m S

ched

ulin

g

require invasive monitoring are more difficult to estimate and almost always take longer than expected. The authors claim that by considering these factors, the ability to predict anesthesia time for a given case could be significantly enhanced27. The ‘anesthesia scheduling packages’ (Table 1) as developed in AMC Amsterdam differentiates in time by taking into account the quantity of monitoring required for the complexity of the procedure as well as for the co-morbidity and thus the complexity of the patient. The package is assigned already during the pre-anesthesia assessment of the patient where medical complexities of the patient related to anesthesia and surgery are evaluated.

Additionally, unanticipated results derived from this study, demonstrating the clinical relevance of this research. These are beneficial side effects due to information that came available earlier in the patient process, which allow for a smoother OR workflow: i.e. the required anesthesia package that is assigned during the pre-anesthesia check-up. Now, anesthesia nurses know exactly which medical equipment and devices need to be assembled and tested beforehand. Correspondingly, anesthesia residents know in advance in which operating room a complex anesthetic technique, like an awake fiberoptic intubation, is scheduled, so they can watch and learn. Moreover, in light of technical skills training, the scheduling and registration of anesthesia packages supplements the clinical training portfolio of residents as well as anesthesiologists.

Furthermore, the difference between the predicted anesthesia package time and the actual realized time per anesthesiologist is a topic during performance review meetings between staff members and the head of the department of Anesthesiology. The main purpose of the department is to improve the quality of anesthesia care. Additional purposes are to create awareness with regard to timeliness and costs, as well as to reduce variation in order to standardize complex work processes.

A final, unanticipated result was the observed improvement in communication between surgeons and anesthesiologists. Because it is now transparent how long anesthesia time will take before the start of every operating room session, surgeons started to take this into account with regard to the complete OR schedule. Nowadays, a patient with a scheduled epidural catheter is placed second on the schedule, instead of first, because anesthesia time will take approximately 60 minutes. Surgeons suggested to start with epidural catheter placement of the second patient while the first patient is still in the OR, which is called ‘parallel processing’. This further supports the idea of Friedman etal.28 that starting the anesthesia process in the preoperative holding area, including intravenous sedation and local anesthesia administration, can realize a significant decrease in the amount of in-room time spent. Moreover, discussing the topic of parallel processing contributes to collaboration and communication in the OR. In a dynamic and complex socio-technical environment, such as an OR, it is important to create awareness and build mutual responsibility and commitment for patients.

However, this study is subject to at least two limitations. Firstly, data were gathered in one

134

6

tertiary referral centre and therefore, general applicability of the findings may be restricted. The mean (SD) total procedure time of 186 minutes (SD 127) reflects that the complexity of procedures is potentially greater than in other hospitals. In a university hospital setting a minimum of 25% up to 30% of total procedure time is engaged by anesthesiologists14. Undoubtedly, this proportion will be smaller in general hospitals. The anesthesia scheduling packages as developed in AMC Amsterdam might be a far too extensive method for other, smaller hospitals organized according to a ‘focused factory’ model, which is characterized by a uniform approach for each patient population segment.

Secondly, dividing OR scheduling into the two main components SCT and ACT, pays no specific attention to the time interval used for positioning, prepping and draping prior to incision. This preparation time is now incorporated in SCT. Especially in a university hospital environment with complex surgical procedures this can take a considerable amount of time and can, therefore, be of influence on the prediction of total procedure time. It would be interesting for future research to focus on differentiating this preparation time per surgical procedures in order to establish whether this would create more accurate predictions.

Finally, descriptive statistics shown in this study demonstrate that especially surgeon-controlled time is subjected to a large variation, indicating that SCT is still the area with the most improvement potential. In AMC, as well as in other Dutch UMCs, efforts have been and are still being made to schedule SCT as realistically as possible by taking into account patient characteristics (such as age, comorbidities, previous hospital admissions, previous surgeries), OR team characteristics (such as years of experience, teaching environment) and procedure characteristics (such as complexity, risk factors, open or laparoscopic). Furthermore, the OR department is used to communicate with surgeons regarding the submitted schedule and to adjust scheduled times if needed, after consulting the specific surgeon. Nevertheless, as aforementioned, SCT is an area with a lot of improvement potential with respect to scheduling. Therefore, AMC recently started to focus on an additional improvement program called “tailor made scheduling”. The main goal of this program is that OR scheduling becomes a combined “team effort” from anesthesiologist and surgeon for each individual, surgical patient. The anesthesiologist makes a realistic estimation by assigning the suitable anesthesia scheduling package and therewith taking into account patient characteristics (such as comorbidities) and the quantity of monitoring required for the complexity of the procedure. The surgeon, on his turn, makes a realistic estimation of the time needed for positioning, prepping, draping and surgery; not merely by using the suggested historical median duration but also by taking into account the specific patient, team and procedure characteristics for every individual case. Realistic scheduling of total procedure time is a shared responsibility of surgeons and anesthesiologists.

By accurately predicting the time required, surgeons and anesthesiologists enable the rest of the team to manage their time and plan for subsequent events to flow smoothly, thus

135

Ane

sthe

sia

Tim

e an

d O

pera

ting

Roo

m S

ched

ulin

g

increasing productivity and improving the harmony of the OR team27. For anesthesia time we recommended new scheduling rules which take into account the anesthetic technique and monitoring required as well as the complexity of the patient. OR scheduling is complex because a procedure entails several elements subject to variability. Scheduling the two major components of a procedure (anesthesia-controlled time as well as surgeon-controlled time) more accurately, and making this a shared responsibility of the professionals involved, leads to less case cancellations, lower prediction errors and smoother OR workflow. This may result in a more efficient use of limited and expensive OR resources.

136

6

REFERENCES

1. Agnoletti V, Buccioli M, Padovani E, et al. Operating room data management: improving efficiency and safety in a surgical block. BMC Surg 2013;13:1-11.

2. Devi SP, Rao KS, Sangeetha SS. Prediction of surgery times and scheduling of operation theaters in ophthalmology department. J Med Syst 2012;36:415-30.




6. Gillespie BM, Chaboyer W, Fairweather N. Factors that influence the expected length of operation: results of a prospective study. BMJ Qual Saf 2012;21:3-12.

7. Lehtonen JM, Torkki P, Peltokorpi A, Moilanen T. Increasing operating room productivity by duration categories and a newsvendor model. Int J Health Care Qual Assur 2013;26:80-92.

8. Pandit JJ, Tavare A. Using mean duration and variation of procedure times to plan a list of surgical operations to fit into the scheduled list time. European Journal of Anaesthesiology 2011;28:493–501.

9. Smith CD, Spackman T, Brommer K, et al. Re-engineering the operating room using variability methodology to improve health care value. J Am Coll Surg 2013;216:559-68.





14. van Veen-Berkx E, Bitter J, Elkhuizen SG, et al. The influence of anesthesia-controlled time on operating room scheduling in Dutch university medical centres

L’influence du temps controle par l’anesthesie sur le programme operatoire dans les centres medicaux d’une universite neerlandaise. Can J Anaesth 2014;61:524-32.



17. van Veen-Berkx E, Elkhuizen SG, van Logten S, et al. Enhancement opportunities in operating room utilization; with a statistical appendix. J Surg Res 2015;194:43-51 e1-2.


137

Ane

sthe

sia

Tim

e an

d O

pera

ting

Roo

m S

ched

ulin

g

19. Dexter F, Coffin S, Tinker JH. Decreases in anesthesia-controlled time cannot permit one additional surgical operation to be reliably scheduled during the workday. Anesth Analg 1995;81:1263-8.


21. van Veen-Berkx E, de Korne DF, Olivier OS, Bal RA, Kazemier G. Benchmarking Operating Room Departments in the Netherlands: Evaluation of a Benchmarking Collaborative between Eight University Medical Centres. Benchmarking: An International Journal 2016;23.

22. Garg R, Bhalotra AR, Bhadoria P, Gupta N, Anand R. Reasons for cancellation of cases on the day of surgery-a prospective study. Indian journal of anaesthesia 2009;53:35-9.

23. Haana V, Sethuraman K, Stephens L, Rosen H, Meara JG. Case cancellations on the day of surgery: an investigation in an Australian paediatric hospital. ANZ J Surg 2009;79:636-40.

24. Kumar R, Gandhi R. Reasons for cancellation of operation on the day of intended surgery in a multidisciplinary 500 bedded hospital. Journal of anaesthesiology, clinical pharmacology 2012;28:66-9.

25. van Veen-Berkx E, Bitter J, Kazemier G, Scheffer GJ, Gooszen HG. Multidisciplinary Teamwork Improves Use of the Operating Room: A Multicenter Study. J Am Coll Surg 2015.

26. Escobar A, Davis EA, Ehrenwerth J, et al. Task analysis of the preincision surgical period: an independent observer-based study of 1558 cases. Anesth Analg 2006;103:922-7.

27. Travis E, Woodhouse S, Tan R, Patel S, Donovan J, Brogan K. Operating theatre time, where does it all go? A prospective observational study. BMJ 2014;349:g7182.

28. Friedman DM, Sokal SM, Chang Y, Berger DL. Increasing operating room efficiency through parallel processing. Ann Surg 2006;243:10-4.

7Preoperative Cross Functional Teams

Improve Operating Room Performance

Justin Bitter, MScElizabeth van Veen-Berkx, MScPierre van Amelsvoort, PhDHein G. Gooszen, MD, PhD

JournalofHealthOrganizationandManagement.2015.29(3):343-52.

140

7

ABSTRACT

Purpose: To present the effect of the introduction of a CFT based organization on planning and performance of Operating Room (OR) teams.

Design/methodology/approach: In this study two surgical departments (A and B) of the Radboud University Nijmegen Medical Centre (RUNMC) in the Netherlands were selected to illustrate the effect on performance. Data was available for a total of seven consecutive years from 2005 until 2012 and consisted of 4,046 OR days for surgical department A and 1,154 OR days for surgical department B on which respectively 8,419 and 5,295 surgical cases were performed. The performance indicator ‘raw utilization’ of the two surgical departments was presented as box-and-whisker plots per year (2005-2011). The relationship between raw utilization (y) and years (x) was analysed with linear regression analysis, to observe if performance changed over time.

Findings: Based on the linear regression analysis, raw utilization of surgical department A showed a statistically significant increase since 2006. The variation in raw utilization reduced from the interquartile range (IQR); IQR 33% in 2005 to IQR 8% in 2011. Surgical department B showed that raw utilization increased since 2005. The variation in raw utilization reduced from IQR 21% in 2005 to IQR 8% in 2011.

Social implications: Hospitals need to improve their productivity and efficiency in response to higher societal demands and rapidly escalating costs. The RUNMC significantly increased their OR performance by introducing a cross-functional team based organization in the operative process and abandoning the so-called ‘functional silos’.

Originality/value: The stepwise reduction of variation in raw utilization – this is shown by a decrease of IQR during the years – indicates an organizational learning effect. This study demonstrates that introducing CFT’s improves OR performance by working together as a team.

141

CFT

s Im

prov

e O

pera

ting

Roo

m P

erfo

rman

ce

INTRODUCTION

Hospitals are continuously exploring ways to simultaneously improve patient safety, quality of care and efficiency. Hospitals need to improve their productivity and efficiency in response to higher societal demands and rapidly escalating costs. In Dutch hospitals most of the growth in costs is being driven by increased health care consumption, partly as a result of medical advances that lead to more diseases at an earlier age so patients can be treated earlier and longer1. Efficient use of OR capacity is crucial since it is considered a high-cost environment and limited hospital resource2. To respond to these challenges, the Radboud University Nijmegen Medical Centre (RUNMC) in the Netherlands opted for a redesign consisting of a cross-functional team (CFT) based organization in the operative process.

In this study we focus on the performance indicator ‘OR utilization’ (for definition: see Methods section). Professionalization of multidisciplinary collaboration and improving planning are ways to improve on this indicator. Mathieu etal.3 show that effective collaboration between professionals is based on attitude, culture and structure (division of tasks and roles). In this study we describe the effect of the introduction of multidisciplinary teams. In our process of ongoing improvement of performance, we decided that once a CFT structure has been implemented and has shown its effectiveness and added value, the focus will be shifted to culture and attitude.

Background theory self-organizing teamsThe roots for the developing theories about CFT’s can be found in the socio-technical systems theory4, 5. Starting point is that organizations have to cope with growing uncertainty and variety. The internal complexity of hospital organizations architecture, caused by traditional functional specialization, is an amplifier for external complexity and a source for interference, errors, variance and accidents. These are difficult to handle due to defects in effective collaboration of autonomous individual professionals. Organizational redesign will revitalize the organization5, 6. Decreasing organizational complexity by reducing the functional concentration and increasing local control is necessary to create optimal conditions for cross-functional teamwork. CFT’s with a high level of self-organizing capabilities and mandate can handle variety, interference and upcoming errors6, 7. Integration of tasks by a cross-functional team-based organization is supposed to reduce the sources of interference, like X-ray equipment not being available or inadequately consulted schedule deviation. Furthermore, cross-functional teams with full mandate are equipped to regulate interference, errors and learn to improve planning under circumstances of scarce resources and large variety.

In this paper we describe the relationship between the implementation of CFT’s and OR performance in the RUNMC in the Netherlands. We have chosen to investigate two surgical departments (A and B) based on their differences in working environment and organizational

142

7

learning effects. Department A is a surgical department performing highly complex, often (sub)acute surgical procedures, frequently demanding intensive care treatment. Department B is a surgical department performing mostly (semi)-elective procedures of mixed complexity, rarely demanding intensive care treatment.

While surgical department B operated in a stable environment during the seven years of investigation, surgical department A went through a turbulent phase of reorganization with strong focus on building a new team with the assignment to improve team performance and patient safety. Surgical department B is a department characterized by a stable – mostly – elective patient population of intermediate and low complexity. On the contrary, surgical department A is characterized by an unstable highly complex and patient population with a large proportion of non-elective procedures. In addition, this population is characterized by a long duration of surgical procedures.

The baselinePrior to redesign, the OR schedule was prepared and controlled by the surgeon in charge. The anesthesiologist approved the schedule the day before. Cancellations regularly occurred due to missing data and other causes, e.g. lack of OR time due to over-utilization at the end of the day. To live up to appointments made with patients, doctor’s commitment also needed improvement. To optimize these pitfalls in the planning and scheduling process, CFT’s were formed in 2004. These CFT’s were called “cross-functional OR scheduling teams” and every surgical department (i.e. orthopedics department, cardiothoracic surgery department etc.) using OR facilities, implemented such a team. The team was supervised by a dedicated anesthesiologist and further consisted of a surgeon, a scheduler, an OR nurse, an anesthesia nurse, a recovery room nurse and a nurse from the ward.

Once a week the team meets to discuss the OR schedule of the following week and to evaluate the OR performance of the previous week, in terms of utilization, cancellations and other factors interfering with smooth planning and performance. The cross-functional team deliberates the complete program, day by day and members inform their colleagues about all relevant issues needed for optimal planning and safety. The CFT is fully responsible for optimal preparation and continuity of the OR program for the week to come. The anesthesiologist as the chairman of the team, chairs the meeting. Besides their role in optimizing OR scheduling, CFT’s draw attention to imminent conflicts.

143

CFT

s Im

prov

e O

pera

ting

Roo

m P

erfo

rman

ce

METHODS

Data was prospectively collected from 2005 until 2012 and analysed retrospectively for the purpose of this study. All data was registered electronically by the OR nursing staff in the Hospital Information System and validated by the surgeon and anesthesiologist in charge. Data used in this longitudinal study involved repeated and continuous measurement of the same performance indicators over a long period of time; in this study raw utilization was focused on. The performance of one OR day, which is generally equal to eight hours of block time (usually from 8:00h until 16:00h) allocated to a specific surgical department, is commonly evaluated by this indicator. It is a measure for efficiency and relates to whether staffed operating rooms are under- or overutilized. An OR is considered underutilized when OR time is staffed but not used for surgery, setup or clean-up, which can occur if cases finish earlier than scheduled, there are prolonged delays between cases or a case is cancelled unexpectedly. An OR is regarded overutilized when it is staffed at overtime wages8.

OR utilization can be calculated in two ways, raw and adjusted. Raw utilization is defined as the total hours of elective cases performed within OR block time divided by the hours of allocated block time per day x 100%. Adjusted utilization uses the total hours of elective cases performed within OR block time, including “credit” for the turnover time necessary to set up and clean up ORs x 100%9-11. This study considered raw utilization, excluding turnover time.

To define a consistent dataset for analysis, all non-elective (emergency) cases and all outpatient cases, were excluded. In RUNMC outpatient surgical cases are allocated to a specific organizational OR unit (a separate ‘day surgery centre’). The outpatient surgery workflow varies from the in-patient surgery workflow. This study focused on elective in-patient surgical cases.

A national independent data management centre was employed to facilitate the collection and processing of the data. This centre provided professional expertise to facilitate the collection and processing of data records. Data reliability checks were performed before data were ready for analysis. Reliability refers to the accuracy and completeness of data, given the intended purpose for use12. Reliability checks for this research consisted of: • A check for missing values (e.g. are all months included; are all OR locations included;

are all required data elements included). • A consistency check to determine if data is in accordance with earlier data deliveries (is

the number of surgical cases comparable with the number of cases during the month of the previous year?).

• The correctness of data was studied to check if values are outside of a designated range (e.g. time patient leaves OR < time patient enters OR; date <> date patient enters OR).

• Outliers were removed from the dataset according to outlier filtering rules (e.g. surgeon-controlled time 0 > x ≤ 1,400 minutes; OR utilization 25 ≥ x ≤ 110%; cumulative turnover time 0 > x ≤ 120 minutes).

144

7

This study evaluated the effect of CFT’s in both surgical departments A and B on OR performance. Surgical department A was chosen because this department went through a rapid and meticulous introduction of this redesign process including optimization of patient safety and surgical scheduling. Moreover, their patient case mix is multifaceted with a relatively high percentage of complex and acute or semi-acute cases, which makes scheduling a demanding process. The Surgical department B was opted for because this is a relatively small group with a high percentage of elective cases.

Data analysis was performed using SPSS Statistics 19. Normality of distribution was determined using the Kolmogorov-Smirnov test. The relationship between raw utilization (y) and years (x), concerning the two surgical departments, was analysed with linear regression analysis. Violations of the basic regression assumptions were diagnosed by means of the residual plot; a graph with the residuals (y - ŷ) plotted on the vertical axe and the predicted values of raw utilization (ŷ) on the horizontal axe.

The box plot, also called a box-and-whisker plot, was introduced by Tukey in 197713. The graphic consists of a box extending from the first quartile (Q1) to the third quartile (Q3); a mark (black horizontal line) at the median; and whiskers extending from the first quartile to the minimum value, and from the third quartile to the maximum value. The interquartile range (IQR) is also called the “middle fifty” and is a measure of dispersion. It is calculated by subtracting the upper and lower quartiles: IQR = Q3 - Q1

14, 15.

RESULTS

Data was available for a total of seven consecutive years from 2005 until 2012. After excluding day care surgery and non-elective surgical cases, the collected data consisted of 4,046 OR days for surgical department A and 1,154 OR days for surgical department B on which respectively 8,419 and 5,295 surgical cases were performed. Outliers (mean ±3 SD) were excluded, based on the SPSS output “Casewise Diagnostics”. This left 4,009 OR days for surgical department A and 1,127 OR days for surgical department B, for statistical analysis.

Data of each year and each department showed that raw utilization was not normally distributed (Kolmogorov-Smirnov test, P< 0.0005). However, normality of data is not an assumption in linear regression analysis. With reference to the basic regression assumptions; interval measure level of variables, independence of the errors, homoscedasticity (or constant variance) of the errors were not violated. Normality of the error distribution was dishonoured, however, this assumption did not lead to biased results because the assumption of normality is not important for large sample sizes (n ≥ 1,000), which was the case in this study.

145

CFT

s Im

prov

e O

pera

ting

Roo

m P

erfo

rman

ce

Figure 1. Box-and-whisker plots raw utilization (%) surgical departments A and B (2005 – 2011)Source:DatabaseDutchORBenchmarkUniversityMedicalCenters,specificallyRUNMC

Surgical department AFigure 1 shows that raw utilization of surgical department A demonstrated an increase since 2006. Most of this increase was effectuated in the lower quartile (Q1 from 62% in 2005 to 91% in 2011) and the median (from 82% in 2005 to 97% in 2011). The variation in raw utilization reduced from IQR 33% in 2005 to IQR 8% in 2011. Results of linear regression analysis showed mean raw utilization significantly increased 3.077% every year (P< 0.0005).

Surgical department BFigure 1 shows that raw utilization of surgical department B demonstrated an increase since 2005. The main part of this increase was effectuated in the lower quartile (Q1 from 74% in 2005 to 86% in 2011). The variation in raw utilization reduced from IQR 21% in 2005 to IQR 8% in 2011. Results of linear regression analysis showed mean raw utilization significantly increased 0.899% every year (P< 0.0005).

146

7

DISCUSSION

The purpose of this study was to identify the relationship between the implementation of cross-functional teams and OR performance. This study shows a long-term perspective and how data illustrate a gradual improvement in OR utilization. One cannot argue with certainty that this is caused by the CFT’s, however, it is not improbable that at least part of the improvement can be ascribed to the systematic work of the two teams. The effects of single loop and double loop learning, as well as focusing on OR scheduling leads to organizational learning4, 7.

Increasing OR performance significantly by the introduction of a CFT based organization in the operative process contributes to more focusing on OR scheduling and an enhanced collaboration as a team. Abandoning the so-called ‘functional silos’ results in less variation in raw utilization. This redesign is based on the principle that a cross-functional team has the ability to attenuate variability, unpredictability and politics6. In other words, cross-functional teams are assumed to have a self-regulating capacity. It is crucial that the OR management facilitates the CFT’s and backs them up, in essence, under any circumstances. Applying socio-technical systems theory design principles has shown to not only lead to improvements in the quality of working life, but can also contribute to an increase in organizational productivity and patient safety as well as better collaboration between professionals4, 6, 7.

We decided that once this structure is in place and has shown to be effective, the attention will be focused on culture and attitude, as the next step. This study showed a significant reduction in variation of raw utilization since the implementation of cross-functional OR scheduling teams in 2004, with a gradual improvement over the years. Regarding to the linear regression analysis, we can conclude a significant increase in mean raw utilization every year, with 3.077% for surgical department A and 0.899% for surgical department B. We expect that this increase will stabilize during the time.

There are two potential explanations for these findings; one is the organizational learning effect and the other is more efficient utilization of OR capacity in the strict sense as a result of focusing on the utilization process by the whole OR organization. The stepwise reduction of variation – a decrease of IQR during the years – indicates an organizational learning effect, whereas an increase of raw utilization, reduction of uncertainty and reliability in scheduling are indicators of more efficient utilization of OR capacity16. This indicates a stable process and positive learning effect in both surgical departments.

The redesign in our study is based on the principle that a CFT has the ability to attenuate variability, reduce unpredictability, the impact of local politics and the effect of personal preference and input of the individual staff members of both departments on the OR schedule. Reduction of uncertainties – by means of optimizing multidisciplinary collaboration – will improve OR scheduling17. In other words, CFT’s are assumed to have a self-regulating capacity to identify bottlenecks and to improve continuity.

147

CFT

s Im

prov

e O

pera

ting

Roo

m P

erfo

rman

ce

The effect of CFT’s can also be endorsed by Donabedian’s traditional structure-process-outcome model18. This model claims a causal relationship between structure, process (CFT’s) and outcome (raw utilization). The structure of the context in which health services are delivered, has an effect on processes and outcomes. Outcomes indicate the combined effects of structure and process.

CFT’s have several indicators to score their performance. Based on previous work of others on improvement of OR scheduling19, 20, raw utilization was chosen in this study since we wanted to analyse over-all performance in a straightforward fashion. CFT’s have shown to progressively learn how to deal with their new role and improve their performance continuously through collaboration and better use of checks and balances4, 6, 21.

Systems that are highly differentiated generally require correspondingly high degrees of integration22-24. As for surgery, accurate scheduling of operations is a crucial factor, complicated by the uncertainty regarding the adequate preparation of the patients on the tentative list and unpredictability of the duration of surgical procedures. Modelling that variability by continuous registration, in turn, provides a mechanism to generate tools for accurate time estimation25. OR professionals are conservative and have a tendency to remain within their comfort zone. Introducing CFT’s is a multi-factorial and multi-consequential intervention with emphasis on multidisciplinary collaboration.

Multidisciplinary teamwork is an important foundation for an effective organization26. Effective CFT’s are characterized by setting and accepting common operational and safety goals3. In effective CFT’s there is a strong collective responsibility for these results in which individual interests are subordinate to the interests of the team3, 26, 27. Effective CFT’s are well organized7 and use single-loop and double-loop learning, as well as feedback processes to continuously learn and improve their performance4, 6.

Gittell27 describes the critical concept of relational coordination. Coordinating work through shared goals, shared knowledge, and mutual respect. Because of the way healthcare is organized, weak links exist throughout the chain of communication. Relational coordination strengthens those weak links, enabling providers to deliver high quality, efficient care to their patients. The result of this study suggest that both the surgical teams have gone successfully through this phase of adaptation to a different planning and control process.

In this study, OR performance was investigated. For reasons explained in the method section, we have chosen to investigate the effect of introduction of CFT’s in the two selected surgical departments A and B based on their differences in case mix, urgency and scheduling challenges. The performance of both departments over the years showed that there is a learning curve and further improvement can be anticipated. Surgical department A showed a stronger organizational learning effect, which was attributed to their unstable relationship to safety, incidents and changes of management over the years. Due to the stable situation of surgical department B a weaker learning effect occurred28.

148

7

The analysis of several additional separate performance indicators, e.g. over-utilized time and case cancellations, can identify areas of further improvement. Other performance indicators – e.g. first-case tardiness, turnover time between cases and under-utilized time – could also further improve utilization of the available OR time. The RUNMC did not specifically formulate goal-settings or standards for OR performance indicators in advance. Even though the Audit Commission29, 30 in the UK has tried to formulate a standard for utilization, a general global standard has not yet been found for performance indicators in OR scheduling. Through benchmarking with other Dutch University Medical Centres, we might be able to substantiate the added value of CFT’s to all other on-going improvement programs.

A limitation of this study was the longitudinal and retrospective nature. During the seven years of investigation other developments parallel to the introduction of CFT’s, e.g. more focus on patient safety issues, and increased awareness of costs and efficiency by national developments in health care, could have influenced the outcome. The dataset of RUNMC was not compared to the other seven University Medical Centres in the Netherlands and no information about their performance is available yet. To further specify the separate role of CFT’s, the data of this study need to be compared to performance data in the other UMC’s in The Netherlands in the near future.

Moreover, there is a difference in patient case mix between the two surgical departments investigated. However, it is unlikely that difference in case mix fully explains the differences in OR performance.

This study that set out to analyse the effects of a new strategy to improve OR efficiency, demonstrates that introducing CFT’s improves OR performance by working together as a team. The results need to be extended and supported by multi-centre research.

149

CFT

s Im

prov

e O

pera

ting

Roo

m P

erfo

rman

ce

REFERENCES

1. Kuenen JW. Provide value. More quality for less money: what the Dutch health care can learn from Sweden. Boston; 2011.


3. Mathieu J, Maynard MT, Rapp T, Gilson L. Team effectiveness 1997-2007: A review of recent advancements and a glimpse into the future. Journal of Management 2008;34:410-76.

4. Argyris C. Single-Loop and Double-Loop Models in Research on Decision-Making. Admin Sci Quart 1976;21:363-75.

5. Sitter LU, Den Hertog JF, Dankbaar B. From Complex Organizations with Simple Jobs to Simple Organizations with Complex Jobs. Human Relations 1997;50.

6. Achterbergh J, Vriens D. Organizations: Social Systems Conducting Experiments. Berlin Heidelberg: Springer-Verlag; 2009.

7. Bitter J, van Veen-Berkx E, Gooszen HG, van Amelsvoort P. Multidisciplinary teamwork is an important issue to healthcare professionals. Team Performance Management 2013;19:263-78.

8. Macario A. Are your operating rooms ‘efficient’? OR Manager 2007;23:16-8.

9. Dexter F, Macario A, Traub RD, Lubarsky DA. Operating room utilization alone is not an accurate metric for the allocation of operating room block time to individual surgeons with low caseloads. Anesthesiology 2003;98:1243-9.

10. Donham RT. Defining measurable OR-PR scheduling, efficiency, and utilization data elements: the Association of Anesthesia Clinical Directors procedural times glossary. International anesthesiology clinics 1998;36:15-29.

11. van Veen-Berkx E, Elkhuizen SG, Kalkman CJ, Buhre WF, Kazemier G. Successful interventions to reduce first-case tardiness in Dutch university medical centers: results of a nationwide operating room benchmark study. Am J Surg 2014;207:949-59.

12. Bowling A. Research Methods in Health. Investigating Health and Health Services. Berkshire: Open University Press 2009.

13. Tukey JW. Exploratory Data Analysis. 1st ed. New Jersey: Pearson PLC; 1977.

14. Dawson R. How Significant Is A Boxplot Outlier? Journal of Statistics Education 2011;19:1-13.

15. Munro BH. Statistical Methods for Healthcare Research. 5th ed. Philadelphia: Lippincott Williams & Wilkins; 2005.



18. Donabedian A. Evaluating the quality of medical care. Milbank Mem Fund Q 1966;44:Suppl:166-206.

19. Stepaniak PS, Heij C, Mannaerts GH, de Quelerij M, de Vries G. Modeling procedure and surgical times for current procedural terminology-anesthesia-surgeon combinations and evaluation in terms of case-duration prediction and operating room efficiency: a multicenter study. Anesth Analg 2009;109:1232-45.


150

7


22. Berg M, Schellekens W, Bergen C. Bridging the quality chasm: integrating professional and organizational approaches to quality. Int J Qual Health Care 2005;17:75-82.

23. Glouberman S, Mintzberg H. Managing the care of health and the cure of disease--Part II: Integration. Health Care Manage Rev 2001;26:70-84; discussion 7-9.

24. Glouberman S, Mintzberg H. Managing the care of health and the cure of disease--Part I: Differentiation. Health Care Manage Rev 2001;26:56-69; discussion 87-9.


26. Parker GM. Cross Functional Teams. Working with allies, enemies and other strangers. San Francisco: Jossey-Bass. An Imprint of Wiley; 2002.

27. Gittell JH. High Performance Healthcare: Using the Power of Relationships to Achieve Quality, Efficiency and Resilience.: The McGraw-Hill Companies; 2009.

28. Pfeffer J, Salancik GR. The External Control of Organizations. A Resource Dependence Perspective. Stanford: Stanford University Press; 2003.

29. Operating Theatres: Review of National Findings. Audit Commission for local authorities and the National Health Service in England & Wales 2003;London.

30. Operating Theatres: a Bulletin for Health Bodies. Audit Commission for local authorities and the National Health Service in England & Wales 2002;London.

151

CFT

s Im

prov

e O

pera

ting

Roo

m P

erfo

rman

ce

8Multidisciplinary Teamwork Improves Use

of the Operating Room: A Multicenter Study

Elizabeth van Veen-Berkx, MScJustin Bitter, MScGeert Kazemier, MD, PhDGert J. Scheffer, MD, PhDHein G. Gooszen, MD, Ph.Dfor the Dutch Operating Room Benchmarking Collaborative

JournalAmericanCollegeofSurgeons.2015.220(6):1070-76.

154

8

ABSTRACT

Background: Poor inter-professional collaboration might negatively influence adequate planning of operative procedures. Interventions capable of improving inter-professional collaboration will positively impact professional practice and healthcare outcomes. Radboud University Medical Center (UMC) redesigned their operating room (OR) scheduling method by implementing cross-functional teams (CFTs). In this center, positive effects of CFTs were already demonstrated in a mono-center study. The recent study aims to confirm these effects by comparing the Radboud data with data of six other, similar centers using a nationwide OR benchmark collaborative.

Study Design: The effect of CFTs was measured by the performance indicator ‘raw utilization’. The Kruskal-Wallis one-way analysis of variance was applied to compare OR performance between all seven centers. The Wilcoxon-Mann-Whitney test was used to determine differences in OR performance between Radboud UMC and the control group.

Results: OR performance differed significantly between all seven centers (P < .0005). Radboud UMC demonstrated the highest median raw utilization of 94%, versus 85% in the control group (P< .0005). Box-and-whisker plots validated the reduced variation during the years, indicating an organizational learning effect. Hence, not only a better performance than the control group but also a gradual improvement of this performance over the years.

Conclusions: This study shows that multidisciplinary collaboration in CFTs during the perioperative phase has a positive influence on OR scheduling and utilization of OR time. Other national databases considering mortality rates further support the idea that introducing CFTs is not only an important condition for improving OR performance, but also for improving quality of care.

155

Team

wor

k Im

prov

es O

pera

ting

Roo

m U

se

INTRODUCTION

Hospitals need to optimize quality of care and the level of patient satisfaction in combination with reduction of costs and improvement of their financial assets. The operating room (OR) facility is the hospital’s largest cost and revenue center, and has a major impact on the performance of the hospital as a whole1-4. Fragmentation of hospital’s activities across departmental working silos (i.e. surgical wards, OR department, anesthesiology department, radiology department) obstructs collaboration and leads to suboptimal use of scarce utilities5, like operating rooms. Therefore, managing the OR is hard due to conflicting priorities and preferences of its stakeholders5, but also because of the scarcity of costly resources.

Accordingly, poor inter-professional collaboration might not only negatively influence the delivery of health services and patient care, but also frustrate adequate planning of operative procedures. Interventions that are capable of improving inter-professional collaboration will have a positive impact on professional practice and healthcare outcomes6,7. In recent years, the introduction of cross-functional teams (CFTs) has received considerable attention8-12 as a result of its capacity to optimize autonomous multidisciplinary team properties to benefit efficiency and performance13-17. Studies on the impact of introduction of CFTs in the OR in particular, are limited.

In 2004, Radboud University Medical Center (Radboud UMC) in Nijmegen, the Netherlands, opted for a redesign of their OR scheduling method, by implementing CFTs. During the following years Radboud UMC increased their OR performance significantly18. These results suggest that introducing a CFT-based organization in the perioperative process improves OR performance.

Although in Radboud UMC the positive effect of CFTs was demonstrated in a preceding mono-center study, these results need to be substantiated in a multi-center comparative study. The aim of this study is to compare the effects established in Radboud UMC with the performance of six other institutions, in order to confirm the influence of the implementation of CFTs on OR performance. Therefore, this study presents data from a nationwide OR benchmark collaborative to compare the data collected in Radboud UMC with those of six other, similar university medical centers (UMCs) in the Netherlands during a time period from 2005 to 2013.

156

8

METHODS

Operating room departments of all eight university hospitals in the Netherlands established a nationwide benchmarking collaborative in 200516-19. The objective of this active collaborative is to improve OR performance by learning from each other through benchmarking data and exchanging best practices.

Measuring OR timeDuring OR sessions, OR nursing staff prospectively registered (in the electronic hospital information system in each UMC) the times for each case occupying the OR. The surgeon and anesthesiologist in charge validated these times after completion of the session.

Each UMC provides their data records for all surgical cases performed to a central OR benchmark database. An independent data management center processes all longitudinal data records and also performs reliability checks preparatory to data analysis. The extensive database is used to calculate key performance indicators related to (non-)utilization of OR capacity and to perform research on OR scheduling issues. The central OR benchmark database, currently comprising 1,279,727 cases in total, consists of records of surgical cases performed at eight UMCs over a nine-year period from 2005 up to and including 2013.

Cross-Functional TeamsRadboud UMC redesigned their OR scheduling method, by implementing the intervention of “cross-functional OR scheduling teams”. Every CFT is headed by a dedicated anesthesiologist and further consists of a surgeon, a scheduler, an OR nurse, an anesthesia nurse, a recovery room nurse and a nurse from the ward. The team meets once a week to discuss the OR schedule of the next week and to evaluate the OR performance of the previous week, in terms of utilization, cancellations and other factors interfering with smooth planning. The CFT examines the complete OR program, day by day and members inform their colleagues regarding all relevant issues needed for optimal planning and safety. The CFT was given full ‘mandate’ (or ‘authorization’) by the Head of the Department of Operating Rooms and by the Head of the Department of Anesthesiology, to make operational decisions regarding the OR schedule and to make alterations to the submitted OR schedule (e.g. change the order of cases or to not approve of a submitted schedule when the scheduled time exceeds the 8h OR block time allocated to a specific surgical department).

The CFT is fully responsible for optimal preparation and continuity of the OR schedule for the upcoming week. The anesthesiologist as the chairman of the team, chairs the meeting. Once the OR schedule of the next week has been accepted, no changes are allowed to be made without the chairman knowing and approving. Besides their role in optimizing OR scheduling, CFTs also pay attention to and discuss imminent conflicts. CFT’s have operational

157

Team

wor

k Im

prov

es O

pera

ting

Roo

m U

se

mandate/authorization to avoid conflicts and to alter OR schedules. Naturally, before alterations are made, they will be discussed with the surgeon responsible. Situations in which discussion with the CFT does not lead to an agreement, the next step will be escalation to the level of the Head of the Department of Operating Theatres and the Head of the Surgical Department involved.

Study population To define a consistent dataset for analysis, we included one and the same specific surgical department in all centers in this study. These departments share their basic logistic challenges and have similar patient populations with roughly the same ratio for elective/emergency cases.

Hence, data from seven UMCs comprising nine consecutive years (2005-2013) were included. A total of 63,607 cases (inpatient elective as well as emergency cases) were subjected to statistical analysis. These 63,607 cases were completed during 30,203 OR-days on which key performance indicators were calculated (per day).

Effect of CFT on utilization as a performance indicatorThis study evaluated the effect on OR performance of a CFT based organization implemented in Radboud UMC, by means of a controlled design, considering empirical OR data of six anonymous control UMCs without this specific CFT based organization. The performance of one OR day, which is generally equal to 8 hours of block time allocated to a specific surgical department, is universally expressed as the indicator “raw utilization”. Raw utilization was defined as the total amount of time patients are present in the OR, divided by the total amount of allocated block time per day x 100%. This definition of raw utilization excluded turnover time and over-utilized OR time19.

Organizational characteristics: case duration and emergency casesOrganizational characteristics concerning total case duration as well as the ratio elective/emergency cases were described. Total case duration (in minutes) was defined as ‘patient in to patient out of the OR room’. In other words, anesthesia-controlled time plus surgeon-controlled time17. The ratio elective/emergency cases was defined as the proportion (%) of elective surgical cases scheduled in advance and the proportion of non-elective/emergency surgical cases performed.

Statistical analysisData analysis was performed using SPSS Statistics 19 (IBM SPSS Statistics for Windows, version 19.0, IBM Corp. Released 2010.; Armonk, NY, USA). Normality of distribution was determined using the Kolmogorov-Smirnov test. Raw utilization was analyzed with the following descriptive statistics: mean (SD), median, inter-quartile range (IQR), and box-and-

158

8

whisker plots. The non-parametric Kruskal-Wallis one-way analysis of variance was applied to compare OR utilization between all seven centers. The non-parametric Wilcoxon-Mann-Whitney test was used to determine differences in OR utilization between Radboud UMC and the six control UMCs together as a group (hereafter referred to as ‘the control group’). To measure the influence of the implementation of new regulations regarding CFTs in May 2012 in Radboud UMC, a (quasi-experimental) time-series design was applied and multiple time periods before and after this intervention were evaluated20. For that reason relevant data concerning mean raw utilization was divided into four equal periods of time. The four different periods in the time-series design were compared with an analysis of variance (ANOVA). To test if the implementation of new regulations regarding CFTs (“the intervention”) led to a significant difference in raw utilization, a contrast analysis was applied: an intervention contrast, a pre-intervention contrast, as well as a post-intervention contrast were tested. Prior thereto Levene’s test was examined. Violations of the basic ANOVA assumptions were examined. The nonparametric alternative to the oneway ANOVA, the Kruskal-Wallis one-way analysis of variance, was used to confirm parametric testing.

RESULTS

General resultsA total of 30,203 OR-days on which the key performance indicator raw utilization (%) was calculated and on which 63,607 inpatient surgical procedures were performed, were selected for inclusion in this study. The organizational characteristics concerning total case duration accompanied by the ratio elective/emergency cases are shown in Table 1 (a; b; c; d). The results of the descriptive statistics of raw utilization are shown in Table 2 (a; b). All numbers are first demonstrated per UMC and second per group, meaning Radboud UMC contrasted with one control group of six anonymous UMCs.

Descriptive statistics demonstrated that, on an overall level, mean case duration ranged from a minimum of 218 (SD 95) minutes to a maximum of 291 (SD 132) minutes. Mean case duration in Radboud UMC was 257 (SD 126) minutes and differed only 12 minutes compared with the control group with a mean case duration of 245 (SD 123) minutes. The ratio elective/emergency cases of 91/9% in Radboud UMC varied from the ratio of 84/15% in the control group.

Effect of CFT on OR performanceThe Kruskal-Wallis one-way analysis of variance test revealed significant differences in OR performance (raw utilization %) between all seven centers (P< .0005). Moreover, this test exposed that Radboud UMC had the highest mean rank (17,617.16) and the highest median

159

Team

wor

k Im

prov

es O

pera

ting

Roo

m U

se

raw utilization (94%, versus 85% group median six centers) during the years 2005 up to and including 2013 (P< .0005). Figure 1 illustrates the difference in raw utilization between Radboud UMC and the control group per year.

Table 1a. Total case duration in minutes (patient in to patient out of the OR) per UMC

Total case duration in minutes

UMC N Mean SD Median 25th percentile 75th percentile

Radboud UMC 10,908 245 123 247 175 303

UMC1 6,081 287 112 282 232 335

UMC2 5,730 218 95 218 160 269

UMC3 6,159 290 142 295 190 375

UMC4 15,091 238 128 238 135 309

UMC5 10,653 242 116 239 179 298

UMC6 8,985 291 132 289 221 357

Table 1b. Total case duration in minutes (patient in to patient out of the OR), Radboud UMC versus control group

Total case duration in minutes

UMC N Mean SD Median 25th percentile 75th percentile

Radboud UMC 10,908 245 123 247 175 303

6 UMCs 52,699 257 126 255 175 324

Table 1c. Absolute number of and ratio elective and emergency cases per UMC

Elective and Emergency cases

Elective cases

Emergency cases

Cases not labelled

Total cases

Elective cases

Emergency cases

Cases not labelled

N N N N % % %

Radboud UMC 9,943 965 0 10,908 91 9 0UMC1 5,185 882 14 6,081 85 15 0UMC2 4,985 745 0 5,730 87 13 0UMC3 5,468 677 14 6,159 89 11 0UMC4 12,552 2,339 200 15,091 83 15 1UMC5 8,948 1,638 67 10,653 84 15 1UMC6 7,314 1,580 91 8,985 81 18 1

160

8

Table 1d. Absolute number of and ratio elective and emergency cases, Radboud UMC versus control group

Elective and Emergency cases

Elective cases

Emergency cases

Cases not labelled

Total cases Elective cases

Emergency cases

Cases not labelled

N N N N % % %

Radboud UMC 9,943 965 0 10,908 91 9 0

6 UMCs 44,452 7,861 386 52,699 84 15 1

Table 2a. Descriptive statistics of raw utilization (%) per UMC

Raw utilization (%)

UMC N Mean SD Median 25th percentile 75th percentile IQR

Radboud UMC 5,300 86 18 94 79 98 19

UMC1 1,549 83 16 89 73 96 23

UMC2 2,373 87 13 91 85 95 10

UMC3 4,103 76 19 81 65 92 27

UMC4 7,462 78 16 82 68 92 24

UMC5 5,031 79 16 83 70 93 23

UMC6 4,385 84 15 89 76 96 20

Table 2b. Descriptive statistics of raw utilization (%), Radboud UMC versus control group

Raw utilization (%)

UMC N Mean SD Median 25th percentile 75th percentile IQR

Radboud UMC 5,300 86 18 94 79 98 19

6 UMCs 24,903 80 17 85 71 94 23

Figure 2 shows box-and-whisker plots per year, displaying the variation of raw utilization in Radboud UMC and the control group. These plots validated the reduced variation (reduction in IQR) in Radboud UMC during the years. The stepwise reduction of variation - a decrease of IQR during the years - indicated an organizational learning effect, whereas an increase of raw utilization, reduction of uncertainty and reliability in scheduling were a display of more effi cient utilization of OR capacity. Hence, not only a better performance in term of OR utilization than the other centers but also a progressive improvement of this performance over the years.

Effect of the implementation of new regulations regarding CFTs on OR performanceRaw utilization did not change significantly when new regulations regarding CFTs were implemented in May 2012 in Radboud UMC (pre-intervention contrast P < .552; intervention

161

Team

wor

k Im

prov

es O

pera

ting

Roo

m U

se

Figure 1. Effect of the introduction of CFTs on OR performance expressed in median raw utilization (%) per year, Radboud UMC versus the control group consisting of six other centers

contrast P < .359; post-intervention contrast P < .894, ANOVA). As a consequence, the implementation of these new regulations had no additional impact on OR performance. Correspondingly, Levene’s test was not significant (P < .695) thus equal variances were assumed between the four equal time periods (two before and two after the intervention).

DISCUSSION

The purpose of this study was to compare the effect of the introduction of CFTs on the OR performance, by comparing data from one university hospital with, with those of six other university hospitals without CFTs. This study demonstrates a gradual improvement in OR utilization in Radboud UMC during the study period. The control group demonstrates a less obvious improvement, compared to Radboud UMC if calculated over time.

The results show that multidisciplinary collaboration in CFTs during the perioperative phase has a positive influence on OR scheduling and utilization of OR time. The initial mono-center findings9 are now supported by the recent findings collected in this multicenter longitudinal study with data of nine consecutive years.

162

8

Figure 2. Effect of the introduction of CFTs on OR performance expressed in box-and-whisker plots of raw utilization (%) per year. Radboud UMC results are compared with those of six other centers in the control group.

An important finding to emerge from the analysis is that a reduction in variation of utilization was observed in Radboud UMC during the years. There are two potential explanations for these findings; one is the organizational learning effect and the other is the result of the focus on the OR scheduling process by the complete OR organization. The assumption is that the outcome of the entire OR process will improve by reducing the variation and uncertainty of multiple elements, like prediction accuracy, availability of specialized OR staff, or availability of discipline- and surgeon-specific equipment21,22. The stepwise reduction of variation - a decrease of the inter-quartile range during the years - indicates an organizational learning effect, while an increase of raw utilization, reduction of uncertainty and reliability in scheduling are indicators of a more efficient utilization of OR time. This indicates a stable process and positive learning effect in the OR department of Radboud UMC.

This study extends earlier observations of improvements by CFTs in the perioperative phase of OR scheduling and confirms previous research on this topic6,9. CFTs are able to improve OR performance by dealing with variability and interferences in the OR process. An organization that is able to achieve a range of objectives, despite variability and interferences, is said to be ‘in control’23.

163

Team

wor

k Im

prov

es O

pera

ting

Roo

m U

se

Importantly, the collaboration in CFTs is able to control the complex scheduling process through single-loop and double-loop learning. CFTs have learned how to deal with interferences, and how to improve their services continuously through better collaboration and better use of control mechanisms (i.e. feedback loops, policy adjustment). An OR schedule, well prepared by a CFT, will reduce the number of cancellations and improve the prediction process for the next schedule. Multidisciplinary collaboration in the perioperative phase contributes to an efficient OR schedule. This is expected to keep waiting lists for ORs as short as possible6.

The findings in this study are subject to at least two limitations. Firstly, the surgical department of Radboud UMC under study showed a strong organizational learning effect with more spread/variation during the first years in this research, which can probably be attributed to a rapidly growing attitude to optimize patient safety in that period of time24. Secondly, it is unclear whether our findings apply to large non-academic hospitals where similar surgery is also performed. Since, in general, the scale is smaller in most of the non-academic hospitals, efficiency may be higher. Therefore a new study focusing on the effects of the difference in organizational characteristics like case duration, prediction accuracy, and patient case mix1 on performance is challenging. To understand the contribution of CFTs, other surgical specialties should also be investigated in a multicenter study to discover whether these findings are similar to the results presented in this study.

Additionally, it is not yet clear whether multidisciplinary collaboration in CFTs also leads to better quality of care6,25,26 since we have not investigated this explicitly in the current study. There are, however, data to support this view since in 2007 The Netherlands Association for Cardiothoracic Surgery, for instance, established the Adult Cardiac Surgery Database24. This dataset comprises demographic factors, type of intervention, in-hospital mortality and 18 risk factors for mortality after cardiac surgery, according to the European System for Cardiac Operative Risk Evaluation definitions. Completeness of data is excellent and national coverage of all 16 Dutch cardiothoracic surgery centers has been achieved since the start. The primary goal of the database is to control and maintain the quality of care by evaluation of outcomes24. According to this database, Radboud UMC has the lowest mortality, and lowest complication rates in comparison with the control group27. The findings in the database of The Netherlands Association for Cardiothoracic Surgery further support the idea and conclusion of our recent study that introducing CFTs is, not only a new and important condition for improving OR performance, but also for improving quality of care. This, however, cannot be concluded from the data presented in this study and should be the topic of future research.

164

8

REFERENCES

1. Cardoen B, Demeulemeester E, Belien J. Operating room planning and scheduling: A literature review. European Journal of Operational Research. 2010;201:921-32.

2. Macario A, Vitez TS, Dunn B. et al. Where are the costs in perioperative care? Analysis of hospital costs and charges for inpatient surgical care. Anesthesiology 1995;83:1138-44.

3. Bodenheimer T, Fernandez A. High and rising Health Care Costs part 4: Can Costs be Controlled While Preserving Quality? Annals of Internal Medicine 2005;143:26-31.

4. Krupka DC, Sandberg WS. Operating room design and its impact on operating room economics. Current opinion in Anaesthesiology 2006;19:185-91.

5. Glouberman S, Mintzberg H. Managing the Care of Health and the Cure of Diseases-Part 1: Differentiation. . Health Care Manage Review 2001;26:56-92.

6. Bitter J, van Veen-Berkx E, Gooszen HG. et al. Multidisciplinary teamwork is an important issue to healthcare professionals. Team Performance Management 2013;19:263-78.

7. Weller JM, Barrow M, Gasquoine S. Interprofessional collaboration among junior doctors and nurses in the hospital setting. Journal of Medical Education 2011;45:478-87.

8. Santa R, Ferrer M, Bretherton P. et al. Contribution of cross-functional teams to the improvement in operational performance. Team Performance Management 2010;16:148-68.

9. Bitter J, van Veen-Berkx E, van Amelsvoort P. et al.Preoperative Cross Functional Teams Improve OR Performance. In: Journal of Health Organization and Management; 2015.

10. Gittell JH. High Performance Healthcare: Using the Power of Relationships to Achieve Quality, Efficiency and Resilience. United States of America: McGraw-Hill Companies; 2009.

11. Wang M-L, Chen W-Y, Lin Y-Y. et al. Structural characteristics, process, and effectiveness of cross-functional teams in hospitals: Testing the I–P–O model. Journal of High Technology Management Research 2010; 21:14-22.

12. Klopper-Kes AHJ, Meerdink N, Wilderom CPM. et al. Effective cooperation influencing performance: a study in Dutch hospitals. International Journal for Quality in Health Care 2011; 23:94-9.

13. Weaver SJ, Rosen MA, DiazGranados D. et al. Does Teamwork Improve Performance in the Operating Room? A Multilevel Evaluation. The Joint Commission Journal on Quality and Patient Safety 2010;36:133-42.

14. Weller J, Boyd M, Cumin D. Teams, tribes and patient safety: overcoming barriers to effective teamwork in healthcare. Postgrad Medical Journal 2014;90:149-54.

15. Deneckere S, Euwema M, Van Herck P, et al. et al. Care pathways lead to better teamwork: Results of a systematic review. . Social Science & Medicine 2012;75:264-8.


17. van Veen-Berkx E, Bitter J, Elkhuizen SG, et al. The influence of anesthesia-controlled time on operating room scheduling in Dutch university medical centres. L’influence du temps controle par l’anesthesie sur le programme operatoire dans les centres medicaux d’une universite neerlandaise. Can J Anaesth 2014;61:524-32.

18. van Veen-Berkx E, Bitter J, Elkhuizen SG, et al. Comment on Research Article Entitled “Variability of Subspecialty-Specific Anesthesia-Controlled Times at Two Academic Institutions” as published in J Med Syst 2014; 38 (11). J Med Syst 2014;38:51.

165

Team

wor

k Im

prov

es O

pera

ting

Roo

m U

se



21. Nave D. How to Compare Six Sigma, Lean and the Theory of Constraints. Quality Progress 2002;73.

22. Achterbergh J, Vriens D. Organizations: social systems conducting experiments. Berlin Heidelberg: Springer-Verlag; 2009.

23. Sitter LUd, den Hertog JF, Dankbaar B. From Comlex Organizations with Simple Jobs to Simple Organizations with Complex Jobs. Human Relations 1997;50:497-534.

24. Siregar S. Safety in cardiac surgery. Enschede: Utrecht University; 2013.

25. Schraagen JM, Schouten T, Smit M, et al. Assessing and improving teamwork in cardiac surgery. Qual Saf Health Care 2010;19:e29.

26. Pronovost PJ, Freischlag JA. Improving Teamwork to Reduce Mortality. JAMA 2010;15:1721-2.

27. Versteegh M. Transparency of Outcome Indicators for Cardio-Thoracic Surgery. 2013 Annual Report (1-6). Utrecht: The Netherlands Association for Cardio-Thoracic Surgery; 2013.

9Multidisciplinary Teamwork is an Important

Issue to Healthcare Professionals

Justin Bitter, MScElizabeth van Veen-Berkx, MScHein G. Gooszen, MD PhDPierre van Amelsvoort, PhD

TeamPerformanceManagement.2013.19(5/6):263-278.

168

9

ABSTRACT

Purpose – The aim of this paper is to describe the factors that contribute to understanding how collaboration improves performance in operating rooms (ORs) after introducing the concept of cross-functional OR scheduling teams. This concept was investigated at Radboud University Nijmegen Medical Center (RUNMC) in the Netherlands and used on an innovative path based on socio-technical systems (STS) principles designed to address non-routine tasks, variety, interferences and errors related to OR scheduling, with the aim of increasing both staff productivity and patient safety.

Design/methodology/approach – The effects of implementing preoperative cross-functional teams in the OR were compared qualitatively. The researcher observed all of the team meetings, available data and documentation, and thirteen semi-structured interviews were performed with team members for collecting additional data.

Findings – In the literature, we found that the theory of socio-technical systems and the fields of groups dynamics and self-managing teams fit the OR setting. We applied six elements of these theories (setting common goals, cohesion, openness, single-loop and double-loop learning, feedback, and control options) to the aspects found in our study. Our qualitative findings revealed that high-performing teams were able to identify bottlenecks in order to improve continuity of care. The cross-functional teams used several performance indicators to gain insight into their own performance. Consequently, through collaboration, these teams were able to minimise interference and therefore learn. Cross-functional teams learned how to address interferences and improve their quality of service through improved collaboration and the improved use of control mechanisms.

Practical implications – This research highlights the importance of team-based approaches and the need to improve collaboration between healthcare professionals.

Originality/value – The paper confirms the value of implementing the socio-technical systems theory to improve collaboration between healthcare professionals. This case study is a valuable contribution, as it focuses on team-based organisation in preparing an OR schedule.

169

Mul

tidis

cipl

inar

y Te

amw

ork

INTRODUCTION

Modern hospitals are confronted with increased uncertainty and variety with respect to organisation and efficiency. Hospitals are internally complex, as they traditionally are functionally specialised with respect to their organisational structure, and many hospitals should redesign their organisation in order to create a more viable structure1, 2. Healthcare is an important social issue, and stakeholders (for example, patients, governments, and insurers) have expectations of latency, throughput, and safety. Therefore, multidisciplinary teamwork is essential for healthcare professionals to improve efficiency and avoid causing unnecessary harm to the patient. However, the principles of socio-technical systems (STS) have not been applied previously to operating room (OR) scheduling in the preoperative phase at this hospital. Furthermore, operating rooms are expensive to the hospital, and capacity should be utilised as much as possible in response to increasing societal demands and rapidly escalating costs. Most of the increase in cost is due to increased health-care consumption3.

In addition, hospitals continuously search for opportunities to improve both productivity and patient safety. For example, Sorbero et al.4 found empirical evidence supporting the relationship between teamwork and patient outcome. Patient quality is the perceived result of the integrated combination of the cure and care processes rather than the sum of the various parts provided by various specialists5, 6.

Teamwork is not a concern for the healthcare field alone. Many industries have recognised the critical role that teamwork plays in effective operation, particularly industries that deal with high-risk, critical safety environments and tasks such as aviation, military operations, and power generation7. Moreover, in industries such as automotive manufacturing, the value of creating high-performance teams has long been recognised8, 9.

In this study, we addressed the complex collaboration between physicians at RUNMC in the Netherlands by studying physicians’ deeply embedded professional differences and how these differences influence the performance in ORs after the ORs were reorganised in 2004. Qualitative research was performed by investigating two OR teams that perform well and two OR teams that performed less well (based on their net utilisation). Because performance with respect to OR scheduling in the preoperative phase is determined by self-managing teams and group dynamics, this study combined the characteristics of these two areas. Our research question was as follows: What elements are important for creating cross-functional teams (CFTs) that can efficiently prepare OR schedules in the preoperative phase?

This paper is organised as follows. The next section describes the baseline situation and provides background information regarding how RUNMC performed before the organisational redesign. The theoretical framework based on the concepts of self-managing teams and group dynamics is also presented. Next, the methodology used to perform the qualitative research is described. Thereafter, the results are presented and discussed. The paper concludes by highlighting some important research gaps that can be addressed in future studies.

170

9

The baselineBefore the operating rooms at RUNMC were reorganised in 2004, a designated surgeon in the assigned medical profession prepared the OR schedule. This schedule was then sent to the anesthesiologist the day before surgery for approval. Adjustments to the schedule were often required due to missing data, a change in the surgeon’s plans, increased surgery time, and last-minute cancellations. To meet the patient’s needs with respect to the date and time of surgery, patient focus needed to be improved by the healthcare staff.

The redesign was entitled “cross-functionalORschedulingteam”, meaning that specialised teams were created for each surgical profession, and these teams would then work together as a department (e.g. the orthopaedics department, cardiothoracic surgery department, etc.) and use the OR facilities. The newly created teams became responsible for the planning, outcome and organisation of the specific OR facilities and their patients. The underlying goal of forming multidisciplinary teams is to break the silo organisation (a silo is a tall, narrow structure, indicating that the organisation was too vertical (hierarchical) organisation) and focus on self-interest. This approach creates reliable planning, better utilisation of resources, balanced workload, and good preparation. The purpose of this collaborative approach is to improve group-based planning and therefore improve utilisation. Improving the reliability of planning will also lead to higher patient satisfaction.

The cross-functional teams consist of an anesthesiologist, a surgeon, a scheduler, an OR nurse, an anesthesia nurse, a recovery room nurse, and a nurse from the specific ward. The anesthesiologist chairs the team meetings. This team composes the OR schedule for the following week and evaluates OR performance from the previous week. The cross-functional team members inform their colleagues of specific preparations for the surgery of that day, and they are involved in the preparation and continuity of the OR program for the following week.

THE THEORETICAL ARGUMENT

The socio-technical systems theoryA cross-functional team‒based organisation was introduced in the operating room (OR) to increase both staff productivity and patient safety. Hospital ORs are high-cost/high-revenue environments, and the facilities are equipped specifically for performing surgical procedures. In this era of rising costs and declining reimbursements, optimising the effectiveness of the operating room suite and maximising throughput10 are essential. Because this facility is usually a hospital’s highest cost and revenue centre11, it has a major impact on the performance of the hospital as a whole. However, managing an operating room is challenging due to conflicting priorities and preferences among its stakeholders5, 6, as well as the scarcity of costly resources. Moreover, healthcare managers must anticipate the increasing demand for surgical services

171

Mul

tidis

cipl

inar

y Te

amw

ork

by our ageing population12. These factors clearly emphasise the need for improved efficiency and the need to adequately plan and schedule procedures.

OR efficiency is defined functionally in terms of the total time the patients are present in the OR divided by the total amount of allocated OR time per 8-hour day (8:00 through 16:00), multiplied by 100. This definition excludes turnover time and over-utilisation of OR time, and OR efficiency is an important factor in determining OR productivity13-16. In other words, OR efficiency can be used to compare what is actually produced or performed with what can ideally be achieved using the same resources (e.g. money, time, labor, etc.). RUNMC followed an innovation path that was based on principles of socio-technical systems (STS) and designed to address non-routine variety, interference, and errors in order to improve productivity and the quality of working life.

The roots of the developing theories regarding cross-functional teams can be found in the STS theory. The STS approach is designed to harness the personal and technical aspects of organisational structures and processes in order to achieve joint optimisation, with a focused emphasis on achieving excellence in terms of both technical performance and the quality of people’s work. The overall goal is to continuously improve performance by setting goals, monitoring and analysing their progress, and identifying and solving problems on a regular basis1, 2, 17.

The starting point in this approach is to recognise that organisations must cope with increasing uncertainty and variety. The internal complexity of a hospital’s organisational architecture stems from traditional functional specialisation, which amplifies external complexity and can serve as a source of interference, errors, variance, and accidents. These factors can be difficult to address due to a lack of effective collaboration between autonomous individual professionals. Redesigning the organisation can often revitalise the organisation, and decreasing organisational complexity by reducing functional concentration and increasing local control will create optimal conditions for cross-functional teamwork1, 2.

OR-based cross-functional teams (CFTs) with high self-organisation capabilities and feasible mandates can cope with variety, interference, and errors more effectively. Integrating tasks using a cross-functional team‒based approach will reduce sources of interference (for example, X-ray equipment being unavailable or scheduling changes that are inadequately discussed among the staff). Furthermore, fully mandated cross-functional teams are equipped to regulate interference and errors, and can learn to improve planning under adverse circumstances such as scarce resources and high variability. The CFT approach has led to an organisational learning effect.

The main goal of this OR redesign was to reduce organisational complexity and the risk of interference by lowering the number of patient transfer points by decreasing functional concentration and increasing local control capabilities. This redesign was necessary in order to create the optimal conditions for collaboration and cross-functional teamwork. Improving

172

9

collaboration between healthcare professionals and applying STS design principles were expected to improve the quality of working life as well as significantly increase organisational productivity and patient safety. Integrating tasks into a cross-functional team‒based organisation can reduce the number of the interference sources. Furthermore, cross-functional OR scheduling teams can cope with local interference and errors and can improve the allocation of scarce resources1, 2. As a result, interference sensibility can be decreased. Interference sensibility is the sum of human errors, patient variation, conflicts of interest among participants, lack of resources, and variations in of procedure times. According to interference sensibility, if interference cannot be controlled at the source, it will escalate and ultimately affect performance.

Nevertheless, collaboration between OR professionals does not come naturally in the Netherlands18. Establishing effective collaboration between professionals is dependent on attitude, culture, and structure. Therefore, RUNMC opted to change the pre-existing structure, culture, and attitude of its OR and staff.

Most hospitals lack the ability to measure whether or not they provide safe patient care. One of the common sources of interference and errors is poor communication between physicians and nurses, who typically interact with each other but not between groups. Similar to the care pathways described by Pronovost etal.19, the goal of redesign intervention is to improve culture and help physicians and nurses learn from their mistakes. In this approach, the principles of highly reliable organisations are applied, with particular attention paid to institutional variables, team variables, and task variables. After the redesign, the hospitals can then reduce unnecessary complexity and variation by standardising the delivery of care and protocols.

In this process, organisational complexity should be reduced by decreasing functional concentration and increasing local control capabilities in order to create the optimum conditions for collaboration and cross-functional teamwork. The intensive collaboration provided by cross-functional teams accelerates the development of routines, thereby reducing interference and facilitating the team’s ability to cope with interference when it arises.

Autonomy and teamworkThe ultimate goal of working together is to establish an effective collaboration20, 21. Autonomy is both an individual and team concept; some researchers stress that teamwork involves a low level of individual autonomy22, whereas others do not rule out the contribution of individual autonomy to effective teamwork23. This attempt to achieve both individual autonomy and a cohesive team can result in tension within a team, creating a paradox24 that can only be resolved by reaching a suitable balance. If team cohesiveness is relatively high, effective collaboration within the team can be maintained, which is essential for effective teamwork25.

173

Mul

tidis

cipl

inar

y Te

amw

ork

Physicians claim and obtain autonomy in designing and executing their work based on their expert authority. However, managers do not necessarily have authority over physicians due to different levels of education. Therefore, it is essential for both physicians and managers to think and act collectively in order to ensure collaboration and achieve organisational improvements. Effective collaboration enables the hospital to deliver services that are both high in quality and cost-effective20, 21.

Single-Loop and Double-Loop LearningArgyris26 distinguishes between single-loop and double-loop learning. Single-loop learning focuses on solving increasingly unilateral changes and the problems that result from those changes. Double-loop learning is closer to the cause of the problem and is based on feedback received with respect to a prior action. Therefore, according to Argyris26, gaining insight into the cause of a problem and finding an effective means to solve that problem are necessary.

The result of combining unilateral professional orientation and far-reaching specialisation is that long-term employee knowledge is only applicable to a limited field of work. As a result, one might (unintentionally) risk creating a specialisation trap. Thus, although the work is initially more secure, a specialisation trap occurs when the professional sees only his own task, and problems are therefore not connected to other tasks to solve them. The employees will be increasingly “condemned” to their own specific expertise and will develop a routine way of working while always dealing with the same category of questions. Because of routine, this quickly becomes a known method, and solutions fail, leading to a “creativity trap”27. Reflection skills are not necessary in this trap and therefore become lost. Without reflection, the learning cycle is not complete, and the CFT will not improve. Self-reflection, self-criticism, and open-mindedness are all neglected, and skills are underdeveloped27, 28. The specialisation trap reduces the employee’s/ professional’s ability to feel responsible for the entire process. Consequently, the feeling of being part of a social partnership is less pronounced, and the effects and benefits of direct action are not seen or felt. Professionals then find the relationship with their immediate colleagues difficult, and their involvement within the organisation can be troublesome. A professional has successfully established certain professional routines that are continuously improved through single-loop learning. However, with establishing non-routine double-loop learning, these routines are removed and professionals are questioned. When these improvements continue to occur, the learning circle is complete28.

Job Demands and Job ControlIn cross-functional OR scheduling teams, job control arises through the development of routines, and this allows employees to deal with interferences. Single-loop and double-loop learning, constructive feedback26, and trust in each other’s qualities are all important aspects of these routines.

174

9

Cross-functional OR scheduling teams are characterised by the fact that their responsibilities are positioned low in the organisation. It is therefore important that they are able to deal with interferences. The Job Demand-Control model of Karasek29 demonstrates why this is so important. This model is based on the psychological demands of the job and the ability of professionals to reflect upon their own work. According to the model, negative and positive health outcomes can be predicted by these two characteristics. Psychological job demands are stressors that are present in working environments that include high pressure, high work pace, and physically and/or mentally demanding work. Management opportunities are closely linked to the worker’s ability to oversee his/her duties and behaviour. A positive outcome (e.g. motivation and active learning behaviour) occurs when the psychological job demands and self-reflecting options are high. According to the model, the most negative health outcome occurs when the psychological job demands are high and when social support and self-reflection options are low.

The members of a CFT must work closely together in order to create an optimal OR schedule. Therefore, it is extremely important to sustain team effectiveness in order to minimise interference and achieve high OR performance. With effective collaboration, the members of the CFTs can achieve common objectives.

Mathieu etal.30 provide a number of characteristics of team effectiveness. In their review spanning a decade of research regarding communication and cohesion within teams, they identified several key points. These key points have a positive effect on the result reached by Mathieu et al. Improvements in the team process can be achieved when employees ask for feedback, discuss errors, and try new methods with the aim of making adjustments and improvements.

Single-loop learning is the only operational adjustment that does not question norms and values. In double-loop learning, the change in norms and values is central to the operational processes in order to continuously improve these processes2, 26. Moreover, interpersonal processes between team members have a large impact on the effectiveness of the entire team31, 32. On the whole, research has demonstrated that constructive feedback has positive effects on the motivation of team members, interpersonal trust, and ultimately the performance of the team. Furthermore, mutual trust and openness within the team are essential, and a collective belief in success has a positive influence on efficiency. Team climate has been shown to affect the attitude and behaviour of the team members, and a feeling of safety within the team can have a large impact on team effectiveness30, 33.

In focusing on the healthcare system, Glouberman and Mintzberg5, 6 identified four quadrants in the healthcare industry: care, cure, control, and community. These four quadrants demonstrate that there are boundaries that limit communication and collaboration between licensed professions and alternative-care providers. In their research, Glouberman and Mintzberg5, 6 found that those kinds of hospitals end up in four entirely separate organisations,

175

Mul

tidis

cipl

inar

y Te

amw

ork

as each part structures itself in an independent way. Setting different goals makes collaborating difficult because of the delicate balance between private and public interests.

One of the most striking challenging in managing a hospital arises when the members of the board attempt to reconcile the goals of the physicians and managers. On one hand, a physician’s primary goal is to treat individual patients in the best possible way. On the other hand, the manager’s primary goal of is to provide continuity for the entire organisation and to deliver high-quality, cost-effective healthcare services to the population. These differences in perspective are a clear source of potential conflict. For a hospital to be manageable, the professional autonomy and organisational position of its physicians are key factors34-36.

Establishing group goals and receiving feedback are inextricably linked in their ability to affect performance37. For example, receiving timely feedback can improve performance and efficiency and result in the establishment of more challenging goals38.

In hospitals, collaboration between professionals is not self-evident34-36. A variety of theories describe common goal setting, control options, cohesion, openness, single-loop and double-loop learning, and feedback as essential variables for improving collaboration. These variables are the starting point for this research. One of the main causes of the redesign was that patient’s surgeries were often cancelled at the last minute. Better planning will lead to strong collective results rather than sub-optimisation and will shift the focus of the team to the patients. Instead of the professional’s agenda taking priority, the patient’s needs are at the centre following the redesign. Improving the scheduling of patients in the OR is based more on control of the work and continuously improving through learning. Therefore, these theoretical foundations and variables were chosen (Figure 1).

Openness

Group dynamics Feedback

Cohesion

Collaboration

Common goal setting

Self-managing teams Single- and double-loop learning

Control options

Figure 1. Theoretical model with operationalisation of core concept of collaboration

176

9

METHODOLOGY

The primary goal of this research is to improve OR performance, and the level of improvement is determined by the level of collaboration39, 40. Here, we performed a qualitative case study to investigate a contemporary phenomenon in a real-life context. In a real-life context, the boundaries between phenomenon and context are not clearly evident, and multiple sources of evidence are used41. From a holistic point of view, the researcher—whose goal is to avoid tunnel vision—uses data triangulation. Data triangulation uses various sources of information in order to increase the validity of the study42. Here, we used participant observation and qualitative content analysis of written and audio-visual documents.

Qualitative research was performed for four cross-functional teams. The Orthopaedics Department and the Oral-Maxillofacial Department were studied because of their consistently high performance over the seven consecutive years that were measured. We also studied two lesser-performing cross-functional teams (the Cardiothoracic Surgery Department and the General Surgery Department), based on their net utilisation performance over the same seven–year period. The goal of the study was to examine how team-based collaboration impacts team effectiveness in a Dutch University Medical Center by studying the effect of implementing preoperative cross-functional teams in the OR.

Using published findings from the literature, the crucial variables were common goal setting37, control options29, cohesion43, openness44, single-loop and double-loop learning2, 26, and feedback2, 26, 38. Thirteen in-depth, semi-structured interviews were conducted with members of the RUNMC cross-functional OR scheduling teams. The key questions were pre-established, and the interview was also conversational, with questions following from previous responses whenever possible. We specifically selected these specialties because of their better or worse performance with respect to net utilisation of the OR facilities during the seven consecutive years. In each team, the respondents consisted of an anesthesiologist, a surgeon, a scheduler, an OR nurse, an anesthesia nurse, and a recovery room nurse; in addition, a nurse from the specific ward was included for the Oral-Maxillofacial Department. The interviews were recorded with the consent of the interviewees. Coding techniques and procedures for developing Grounded Theory were performed for data processing and reduction of raw data. The Grounded Theory approach is a systematic methodology used the social sciences for the discovery of theory through the analysis of data. This theory is used primarily in qualitative research studies45, 46.

The data were analysed in the following three steps: open coding, axial coding, and selective coding. Strauss and Corbin45 described open coding as “breaking down, examining, comparing, conceptualization and categorizing’’ data. The starting point in this phase is the research material. Codes are a summary format for a piece of text, in which the meaning of the fragment is expressed, is highlighted and given a summary name under which it is stored. Axial coding refers to “a set of procedures whereby data are put back together in new ways

177

Mul

tidis

cipl

inar

y Te

amw

ork

after open coding, by making connections between categories”45. The first aim of axial coding is to identify the major and minor elements of the study. The second aim is to reduce the size of the data and the number of codes. Axial coding is used to organise the codes obtained from the first stage. Axial coding reduces the number of concepts and relates the concepts hierarchically. In selective coding, the goal is “selecting the core category, systematically relating it to other categories, and filling in categories that need further top refinement and development”45. After unravelling the data, the researcher combines and structures the data. In the selective coding phase, the emphasis is on integrating the data and linking the categories.

The authors of this paper analysed and interpreted the data after assigning the key concepts of the study into dimension groups. For each dimension, the results of the study documentation, the respondents’ answers, and the observations obtained from the consultation were compared, resulting in a description of the actual state of collaboration45.

Collaboration was investigated as an independent variable and was operationalised in variables such as single-loop and double-loop learning and feedback2, 26, openness, common goal setting and cohesion30, 43, and control options29. We chose this method in order to explore the qualitative nuances in these relationships, as these relationships could not be analysed using quantitative research methods. We questioned our respondents regarding the way the members of the cross-functional team perform with respect to collaboration, as well as how collaboration positively influences the way in which they perceive the relationship between collaboration and OR performance.

RESULTS

The important foundations of a CFT are to establish common goals, achieve job control (mandate), and apply single-loop and double-loop learning. These three variables differed between the well-performing CFTs and the CFTs that performed less well. The other three variables (openness, cohesion, and feedback) differed to a lesser extent than the first three variables. In the well-performing teams, openness (the atmosphere and collaboration) was rarely the subject of discussion, even though this topic should be discussed regularly44. The well-performing teams showed cohesion, but conflicts arose when the team members were unable to reach a consensus regarding an issue. In this regard, relationships between the team members are critical47. Most conflicts arise when one or more team member is not properly informed.

Although feedback was given in the well-performing teams, it was not always given directly to the person involved. Retrospective feedback was aimed at the process rather than the person. Nevertheless, receiving timely feedback can lead to improved performance, higher efficiency, and establishing more challenging goals38.

After analysing the interviews, documents, and observations, a distinction was made

178

9

between the well-performing CFTs and the teams that performed less well. This distinction was based on the presence of a learning curve during seven consecutive years and the maximum net OR utilisation. Data were collected from January 1, 2005 through December 31, 2011. All data were registered electronically in the Hospital Information System by the OR nursing staff and validated by the surgeon and anesthesiologist in charge. Table 1 summarises the outcomes of the semi-structured interviews, documentation, and observations.

Table 1. Outcome of analyzed interviews, documentation and observations

Inde pendent variable Well-performing cross-functional teams Lesser-performing cross-functional teams

Common goal setting • Patient is central rather than self-interest

• Clear focus on common result

• Different policy principles• Different insight and understanding of

work organisation

Cohesion • A strong sense of shared responsibility

• Participants demonstrate understanding of each position

• Collaboration is organised in a healthcare chain

• Tension between the participants because of their own interests

• Participants do not always show up for meetings

• Professional puts pressure on proposed OR schedule

• Collaboration in silos

Openness • The atmosphere and collaboration are not often the subject of discussion

• Limited policy dialogue underlying insights

• The true discussion is regularly evaded

Single-loop learning • Weekly evaluation provides improvements

• Planning horizon for two weeks is introduced

• Insufficient uniformity for performance indicators (i.e. definition of turnover time)

• Planning not prepared well• Many last-minute repairs necessary

regarding OR schedule

Double-loop learning • Thinking is multidisciplinary• Policy meetings quarterly• Continuous learning• Doubt regarding norms and values

• Thinking in links• No policy meetings• Learning cycle is not complete

Feedback • Direct feedback during meetings and evaluations

• Retrospective feedback is aimed at the process, not to the person involved

• Little or no insight on performance• Little or no feedback (appreciation)• Learning cycle is not complete

Internal control options • Tension between the financial incentive to maximize utilisation versus the workload for staff

• No direct consequences for participants involved

• Little guidance on planning deviation

External control options • External control options are present but constrained by budgets and patient flow

• Insufficient cross-examination collaboration

179

Mul

tidis

cipl

inar

y Te

amw

ork

The interviews yielded the following results for the six variables:(1) Common goal setting:A shared goal allows employees to focus more on the overall

results. If a difference in opinion arises between employees due to self-interest, patient focus can be disrupted. Therefore, understanding each other’s role is important in order to foster mutual respect during the decision-making process.

(2) Cohesion:The well-performing teams arranged suitable replacements in the event of an absence. For example, replacing a permanent team member during a holiday can give rise to conflict if the replacement is not properly informed of the established procedures. Each team member has respect for other members’ opinions, and they usually view a topic from a distance before reacting. Issues must be taken seriously, but cohesion can be lost if no suitable solution or consensus can be reached. The loss of cohesion is less present in non-controllable factors surrounding the planning.

(3) Openness: In the teams that perform less well, the true discussion was regularly avoided, and improvements took longer and did not always create the desired efficiency. Interviewees indicated that openness to discuss the issues and openness with each other are necessary in order to create a pleasant and safe atmosphere. In the well-performing teams, atmosphere and collaboration were not often the subject of discussion.

(4) Single-loopanddouble-loop learning:One of the differences in performance between the two sets of teams was the presence of double-loop learning. In single-loop learning, the CFT members modify their actions according to the difference between expected and obtained outcomes. In double-loop learning, the members of the CFT question the values, assumptions, and policies that led to the actions in the first place; if they are able to view and modify their actions, double-loop learning has taken place, and the transformation from input to output will be improved.

(5) Feedback: Interviewees indicated that feedback is neutral and always coloured by the person who gives the feedback (his/her norms, values , and self-image) and by the relationship between the giver and receiver of the feedback. The more personal the relationship, the greater the likelihood that the feedback receiver will accept the feedback. On the other hand, if the feedback giver has given valuable hurtful feedback, this will likely lead to either an improvement or worsening of the relationship between the two parties.

(6) Controloptions:A CFT with control options (mandate) can handle interferences and solve problems more easily. Because the CFT is positioned low within the organisation (near the workplace), they have access to insights into making improvements. A CFT can arrange their work schedule and their responsibilities for themselves, but they are constrained by budgetary limits and patient flow. With the CFT, decisions are made in order to ensure quality patient care, both internally and externally. Within the CFT, information is processed independently by the members in order to reach good decisions. The CFT has also external control options, such as the ability to temporarily increase OR capacity.

180

9

DISCUSSION

The goal of cross-functional OR scheduling teams is to ensure that the scheduled patients receive the surgery according to the weekly schedule after it is established. Collaboration can reveal conflicting interests, and working together can be complicated by personal communication barriers. By creating a cross-functional OR scheduling team, the interests of the team can become apparent much more quickly. However, collaboration between professionals is not always guaranteed. The organisation of the team meetings with respect to the attitude and behaviour of the team members are key factors for achieving success. Standardisation and establishing protocols can help the team prepare the OR schedule.

In the literature, we found that the fields of group dynamics and self-managing teams fit the OR setting. We applied six elements of the fields of group dynamics and self-managing teams to the aspects found in our study. This led to the following recommendations, which can facilitate effective collaboration in order to help CFTs efficiently prepare the OR schedule in the preoperative phase:

• Address common goals for a collective focus towards reaching a common result.• Arrange control options (mandate) for decision-making at the lowest possible level as

close as possible to where the outcome is realised.• Single-loop and double-loop learning: provide a weekly evaluation of the work (single-

loop learning) and periodically question the norms and values and improve as needed (double-loop learning).

• Create an environment of openness and cohesion in which people can held each other accountable regularly, and in which everyone can contribute something in a safe and receptive environment.

• Create a safe environment in which feedback between the CFT members is constructive and neutral. Timing of this feedback is also important.

Although the results presented in Table 1 are self-explanatory, one of the main reasons underlying the differences in OR performance is the extent to which scheduling uncertainty and reliability are reduced. These factors are relevant for collaboration. The research revealed that the key differences between well-performing teams and less well-performing teams are common goal setting and single-loop and double-loop learning, which are essential for continuous improvement. In particular, double-loop learning and control mandates were important in the higher-performing teams, which were able to accommodate multidisciplinary professions and therefore improved continuously during the study period. Cohesion, openness and feedback are indirectly essential to improving performance. The less well- performing teams did not hold their members accountable for their actions, and the learning circle was not complete. Showing understanding of each other’s role was also lacking.

181

Mul

tidis

cipl

inar

y Te

amw

ork

The team members’ self-interest regularly took precedence over the public’s interest in the less well-performing cross-functional teams. When this happens, the team’s chairperson must intervene to prevent this undesirable behaviour, and the participants themselves must be critical of one another and give constructive feedback.

The cross-functional teams act primarily as well-informed, professional organisations, although frustrations remain and must be addressed. The participants are given the opportunity to be honest and have discussions regarding the organisation, processes, attitudes, and behavior in a safe environment. In the less well-performing teams, these factors could have been improved by creating a better partnership, a fruitful dialogue, increased job control, and more effective conflict handling.

The results of this qualitative research revealed that the best-performing teams could identify bottlenecks in order to improve continuity and productivity. The CFTs gained insight into their performance using several performance indicators. Consequently, through collaboration, a cross-functional team can both control and learn. Cross-functional teams learn how to address interferences and can continuously improve their services through better collaboration and by using control mechanisms more effectively.

This research revealed that implementing cross-functional OR scheduling teams directly improves OR performance. Proactively preparing the preoperative processes through teamwork yields a better outcome on the day of the surgery due to less interference.

As a result of traditional functional specialisation, the internal complexity of a hospital’s organisational architecture is an amplifier of external complexity and a source of interference, errors, variability, and accidents. These complications are difficult to handle, due to the lack of effective collaboration between autonomous individual professionals. This behavior and characteristics can be changed in a complex organisation by creating a multidisciplinary team with double-loop learning, mandates, and the establishment of a common goal. CFTs are responsible for the planning, results, and organisation of the specific OR facilities and its patients. To establish a common goal, the board of directors must formulate a clear objective.

With its socio-technical design, a hospital’s cross-functional OR scheduling team is better prepared to address over-utilisation, under-utilisation, and schedule deviations and, thereby preventing cancellations. With higher employee satisfaction and an increase in the number of patients administered, the facility’s scarce resources can be optimally utilised. Consequently, control options play an essential role in collaboration within a cross-functional OR schedule team.

Collaboration yields a single-loop learning effect. By providing feedback with respect to organisation, processes, attitudes, and behaviour, the cross-functional team can learn from previous experiences and therefore improve continuously. In policy meetings, a double-loop effect is achieved by discussing norms and values and adjusting as needed.

182

9

In this study, collaboration within four cross-functional teams was investigated. The teams were chosen based on the performance of the surgical service, measured as net utilisation. Using performance indicators of net utilisation is likely not the only explanation for the results obtained. For example, the planning horizon, the composition of non-investigated teams, and other trusted variables were not included in this research.

The overall performance of a surgical service can be affected by multiple variables, including the mixture of patient cases, the scarcity of resources, and the OR’s planning horizon. These variables can be investigated in future studies. In addition, performing a study similar to this in other medical centres with the Netherlands will allow comparisons and support the initial conclusions of our study.

This research focused on the organisational process, not the quality of the medical care itself. Although the less well-performing cross-functional teams were well-organised in some respects, in order to improve continuously, these teams should focus on what improvements can be made in the near future.

The outcome of this new strategy to improve OR efficiency demonstrates that introducing CFTs can improve OR performance by allowing the individual healthcare workers to function as a team. Although this study is preliminary, it can serve as a starting point for more comprehensive studies to expand these initial findings.

183

Mul

tidis

cipl

inar

y Te

amw

ork

REFERENCES

1. Sitter LU, Den Hertog JF, Dankbaar B. From Complex Organizations with Simple Jobs to Simple Organizations with Complex Jobs. Human Relations 1997;50.

2. Achterbergh J, Vriens D. Organizations: Social Systems Conducting Experiments. Berlin Heidelberg: Springer-Verlag; 2009.

3. Kuenen JW. Provide value. More quality for less money: what the Dutch health care can learn from Sweden. Boston; 2011.

4. Sorbero E, Donna O, Farley DO, Mattke S, Lovejoy S. Outcome Measures for Effective Teamwork in Inpatient Care. Santa Monica: RAND Corporation; 2008.

5. Glouberman S, Mintzberg H. Managing the care of health and the cure of disease--Part II: Integration. Health Care Manage Rev 2001;26:70-84; discussion 7-9.

6. Glouberman S, Mintzberg H. Managing the care of health and the cure of disease--Part I: Differentiation. Health Care Manage Rev 2001;26:56-69; discussion 87-9.

7. Kozlowski SWJ, Ilgen DR. Enhancing the effectiveness of work groups and teams. Psychological Science in the Public Interest 2006;7:77-124.

8. Cohen SG, Bailey DE. What makes teams work: group effectiveness research from the shop floor to the executive suite. Journal of Management 1997:239-90.

9. Salas E, DiazGranados D, Weaver SJ, King H. Does Team Training Work? Principles for Health Care. Academic Emergency Medicine 2008;15:1002-9.

10. Krupka DC, Sandberg WS. Operating room design and its impact on operating room economics. Curr Opin Anaesthesiol 2006;19:185-91.

11. Macario A, Vitez TS, Dunn B, McDonald T. Where are the costs in perioperative care? Analysis of hospital costs and charges for inpatient surgical care. Anesthesiology 1995;83:1138-44.

12. Etzioni DA, Liu JH, Maggard MA, Ko CY. The aging population and its impact on the surgery workforce. Ann Surg 2003;238:170-7.


14. Strum DP, Vargas LG, May JH. Surgical subspecialty block utilization and capacity planning: a minimal cost analysis model. Anesthesiology 1999;90:1176-85.

15. Dexter F, Epstein RH, Marsh HM. A statistical analysis of weekday operating room anesthesia group staffing costs at nine independently managed surgical suites. Anesth Analg 2001;92:1493-8.

16. Dexter F, Traub RD. How to schedule elective surgical cases into specific operating rooms to maximize the efficiency of use of operating room time. Anesth Analg 2002;94:933-42, table of contents.

17. Cherns A. Principles of Sociotechnical Design. Human Relations 1976;29:783-92.

18. Klopper-Kes AH, Meerdink N, Wilderom CP, Van Harten WH. Effective cooperation influencing performance: a study in Dutch hospitals. Int J Qual Health Care 2011;23:94-9.

19. Pronovost PJ, Berenholtz SM, Goeschel CA, et al. Creating high reliability in health care organizations. Health Serv Res 2006;41:1599-617.

20. Shortell SM, Marsteller JA, Lin M, et al. The role of perceived team effectiveness in improving chronic illness care. Med Care 2004;42:1040-8.

184

9

21. Smalarz A. Physician group cultural dimensions and quality performance indicators: not all is equal. Health Care Manage Rev 2006;31:179-87.

22. Wellins RS, Wilson R, Katz AJ, Laughlin P, Day CR, Price D. Self-directed teams: A study of current practice. Pittsburgh: DDI; 1990.

23. Cohen SG, Ledford GE. The Effectiveness of Self-Managing Teams: A Quasi-Experiment. Human Relations 1994;47:13-43.

24. Manz C. The Illusion of Self-Management: Using Teams to Disempower. New York: Sharpe; 2003.

25. Langfred CW. The paradox of self-management: Individual and group autonomy in work groups. J Organ Behav 2000;21:563-85.

26. Argyris C. Single-Loop and Double-Loop Models in Research on Decision-Making. Admin Sci Quart 1976;21:363-75.

27. Van Delden PJ. Proffeionals: kwaliteit van het beroep. Amsterdam: Veen; 1991.

28. van Amelsvoort P. Als het hek van de dam is, lopen de schapen overal. Nijmegen: Radboud Universiteit Nijmegen; 2007.

29. Karasek RA. Job Demands, Job Decision Latitude, and Mental Strain - Implications for Job Redesign. Admin Sci Quart 1979;24:285-308.

30. Mathieu J, Maynard MT, Rapp T, Gilson L. Team effectiveness 1997-2007: A review of recent advancements and a glimpse into the future. Journal of Management 2008;34:410-76.

31. Jehn KA, Northcraft GB, Neale MA. Why differences make a difference: A field study of diversity, conflict, and performance in workgroups. Admin Sci Quart 1999;44:741-63.

32. De Dreu CK, Weingart LR. Task versus relationship conflict, team performance, and team member satisfaction: a meta-analysis. J Appl Psychol 2003;88:741-9.

33. Edmonson A. Psychological safety and learning behavior in work teams. Admin Sci Quart 1999;44:350.

34. Davies HT, Hodges CL, Rundall TG. Views of doctors and managers on the doctor-manager relationship in the NHS. BMJ 2003;326:626-8.

35. Edwards N. Doctors and managers: poor relationships may be damaging patients-what can be done? Qual Saf Health Care 2003;12 Suppl 1:i21-4.

36. Kaissi A. Manager-physician relationships: an organizational theory perspective. Health Care Manag (Frederick) 2005;24:165-76.

37. Locke EA, Latham GP. Building a practically useful theory of goal setting and task motivation. A 35-year odyssey. Am Psychol 2002;57:705-17.

38. Stansfield TC, Longenecker CO. The effects of goal setting and feedback on manufacturing productivity: a field experiment. International Journal of Productivity and Performance Management 2006;55:346-58.

39. Delarue A, Van Hootegem G, Procter S, Burridge M. Teamworking and organizational performance: A review of survey-based research. Int J Manag Rev 2008;10:127-48.

40. Santa R, Ferrer M, Bretherton P, Hyland P. Contribution of cross-functional teams to the improvement in operational performance. Team Performance Management 2010;16:148-68.

41. Verschuren PJM. Case study as a research strategy: some ambiguities and opportunities. International Journal of Social Research Methodology 2003;6:121-39.

42. Thurmond VA. The point of triangulation. J Nurs Scholarsh 2001;33:253-8.

43. Zaccaro SJ, Lowe CA. Cohesiveness and Performance on an Additive Task - Evidence for Multidimensionality. J Soc Psychol 1988;128:547-58.

185

Mul

tidis

cipl

inar

y Te

amw

ork

44. Hobman EV, Bordia P, Gallois C. Perceived dissimilarity and work group involvement - The moderating effects of group openness to diversity. Group Organ Manage 2004;29:560-87.

45. Strauss AL, Corbin J. Basics of qualitative research: techniques and procedures for Developing Grounded Theory.: Sage Publications, ltd.; 1998.

46. Boyatzis RE. Thematic analysis: coding as a process for transforming qualitative information. Thousand Oaks, CA: Sage Publications Inc.; 1998.

47. Beal DJ, Cohen RR, Burke MJ, McLendon CL. Cohesion and performance in groups: A meta-analytic clarification of construct relations. Journal of Applied Psychology 2003;88:989-1004.

10Dedicated Operating Room for Emergency Surgery

Generates More Utilization, Less Overtime

and Less Cancellations

Elizabeth van Veen-Berkx, MScSylvia G. Elkhuizen, PhDBart Kuijper, MScGeert Kazemier, MD, PhDfor the Dutch Operating Room Benchmarking Collaborative

AmericanJournalofSurgery.2015.211:122-8

188

10

ABSTRACT

Background: Two approaches prevail for reserving OR capacity for emergency surgery: (1) dedicated emergency ORs and (2) evenly allocating capacity to all elective ORs, hereby creating a virtual emergency team. Previous studies contradict which approach leads to the best performance in OR utilization.

Methods: Quasi-experimental controlled time-series design with empirical data of three centers. Four different time periods were compared with analysis of variance (ANOVA) with contrasts.

Results: Performance was measured based on 467,522 surgical cases. After closing the dedicated emergency OR, utilization slightly increased; overtime also increased. This was in contrast to earlier simulated results. The two control centers, maintaining a dedicated emergency OR, showed a higher increase in utilization and a decrease in overtime, along with a smaller ratio of case cancellations due to emergency surgery.

Conclusions: This study shows that in daily practice a dedicated emergency OR is the preferred approach in performance terms regarding utilization, overtime and case cancellations.

189

Ope

ratin

g R

oom

Cap

acity

for E

mer

genc

y S

urge

ry

INTRODUCTION

Operating rooms (ORs) are of paramount importance to a hospital, given the fact that more than 60% of admitted patients are treated in the OR1. Efficient use of OR capacity is pivotal as it is considered a high-cost environment and a limited hospital resource2. Due to the aging population and developments in surgery, demands for OR facilities are likely to increase. Moreover, due to shortages of qualified OR staff, utilization of ORs is an ever-increasing challenge1. For this reason, optimal scheduling of ORs to ensure effective and efficient use of OR capacity is crucial. However, variability in processes, dependence on limited capacity in other parts of the hospital such as intensive care units, large numbers of surgical departments competing for limited OR facilities, and particularly the unpredictable arrival and duration of emergency surgeries render scheduling complex1, 3. Emergency procedures hamper the elective OR schedule, resulting in delays, case cancellations or overtime3.

Previous studies have described different approaches to deal with emergency procedures and the disturbances they create in elective OR schedules3-12. Overall, these different approaches can be divided into two basic methods for reserving OR capacity for emergency surgery: (1) concentrating all reserved OR capacity in dedicated emergency ORs and (2) evenly allocating capacity for emergency surgery to all elective ORs, hereby creating a virtual emergency team. Several studies have suggested to favor approach (1)3, 4, 7-10, 13, while other studies promote approach (2)5, 6, 11, 12. Conclusions of these previous studies contradict with regard to the allocation method leading to the best performance in OR utilization. Many hospitals debate on this subject and in practice ‘closing the dedicated emergency OR’ is becoming the preferred approach.

In 2007 Wullink etal.12 compared the two basic approaches for reserving OR capacity for emergency patients by using a discrete event simulation model simulating the actual situation. Results that were based on a large database indicated that the policy of reserving capacity for emergency surgery in all elective ORs could lead to an improvement in waiting times for emergency surgery from 74 (±4.4) minutes to 8 (±0.5) min, while working in overtime was reduced by 20%, and overall OR utilization increased by 3%. The results of this simulation study led to the closing of the emergency OR at Erasmus University Medical Center Rotterdam (Erasmus MC) and to planning emergency procedures during the day in the reserved slack time of all elective ORs.

A systematic review conducted by Fone etal.14 concluded simulation modeling to be a powerful method to inform policy makers in the provision of health care. Although the number of modeling papers has grown substantially over recent years, few report on the outcomes of implementation of models, therefore the true value of modeling cannot be assessed. It is likely that many modeling studies are published before implementation of the relevant intervention(s) has been carried out and evaluated. OR scheduling is one of the popular topics

190

10

for computer simulation modeling in health care. The potential of mathematical modeling to inform evidence-based policy development in health care is clear, however, information on the outcomes of model implementation and hence the value of modeling requires further research14.

The purpose of this study was to evaluate the policy outcomes of reserving capacity for emergency surgery in all elective ORs, in terms of OR utilization. This policy was assessed using a controlled time-series design and empirical OR data of three University Medical Centers (UMCs) in the Netherlands.

METHODS

The definition of emergency surgery was based on the unplanned nature of identification of the need for surgery and the relative urgency for surgical intervention, without which the patient’s health may deteriorate and risk poor clinical outcomes15. This included all non-elective (unplanned) surgical cases, both urgent and semi-urgent. The timeframe for indication of urgency included all non-elective cases requiring surgery immediately, within two hours, within six hours or within 24 hours.

To measure the influence of the emergency scheduling approach, a quasi-experimental controlled time-series design was applied. Data were retrieved directly from the Hospital Information Systems of three UMCs. The analysis concerned data for the period of eight consecutive years from 1 January 2000 through 31 December 2007. In addition, more recent data from 2008 to 2011 were included in the results, however, these recent years were not part of the actual study period. In this study data from three university hospitals were included: the Erasmus MC which applied a new method for emergency planning, and two control UMCs. These control UMCs were selected, based on comparability with the Erasmus MC in size and patient mix. Three performance indicators were relevant for the evaluation of the emergency planning approach:

• Raw utilization was defined as the total amount of case durations (elective and emergency cases) during block time, divided by the total amount of allocated block time for the complete OR department x 100%. This definition excluded turnover time and overtime. Block time or ‘opening hours’ are generally from 8:00 AM - 16:00 PM. The common scheduled start and finish times were corrected in case of an intentionally alternation, e.g. due to regular team meetings, extended block time.

• The number of operating rooms running after scheduled room exit time (generally 16:00h), divided by the total number of available, staffed ORs x 100%.

191

Ope

ratin

g R

oom

Cap

acity

for E

mer

genc

y S

urge

ry

• Cumulative ‘overtime’ from all operating rooms running late, divided by the total amount of allocated block time x 100%. Overtime was quantified by the difference in minutes between the scheduled and actual room exit time of the last patient of the day. The common finish time was corrected in case of a scheduled extended block time (more than the standard of eight hours).

These parameters were measured daily and prospectively for the OR departments of three UMCs. We divided the available data in four time periods of two years each. With two measurements periods before the implementation (2000/2001 & 2002/2003) and two measurements periods after the implementation (2004/2005 & 2006/2007) we created an interrupted time series design that allowed to control for changes in the parameters not caused by the intervention16. Data analysis was performed with SPSS Statistics 20 (IBM SPSS Statistics for Windows, version 20.0, IBM Corp. Released 2011.; Armonk, NY, USA). The four different periods in the time-series design were compared with an analysis of variance (ANOVA) with contrast analysis. To test whether parameter changes were initiated by the intervention, three contrasts were calculated for each analysis: an intervention contrast, a pre-intervention contrast and a post-intervention contrast. Prior thereto Levene’s Test was examined. Violations of the basic ANOVA assumptions were examined.

In addition, the ratio of the number of cancellations for which the recorded reason was ‘due to an emergency case’ divided by the total number of cancellations, was calculated for the three centers. A cancellation on the day of intended surgery was defined as an operation that was scheduled on the final OR schedule for that day (generated at 15:00h the day before) and that was not performed on that day17. Each cancellation with associated reasons for cancellation was registered in the Hospital Information Systems of the three UMCs.

RESULTS

The performance indicators were measured based on a total amount of 467,522 surgical cases in the three university hospitals together; a mean of 155,841 cases per hospital. Figure 1 shows the mean raw utilization (%) of the OR departments of the three UMCs during and after the study period. In the Erasmus MC utilization slightly increased after the intervention in 2003. However, in the two control UMCs, positive changes were visible as well. To test whether changes were significant, an ANOVA-contrast analysis for four sequential periods was conducted.

Table 1 summarizes the parameter values for the three UMCs during the different measurement periods. Table 2 illustrates the results of the ANOVA-contrast analysis. To attribute a difference in performance to the intervention, the expectation is that for the Erasmus MC the intervention contrast was significant (P < 0.01) and both the pre- and post- measurement

192

10

contrasts were not. This was the case regarding raw utilization (%). The value of the contrast was positive meaning that utilization was higher after the intervention than before.

However, both control UMCs also showed significant differences between the measurement periods. UMC1 showed significant results for all contrast analyses (P < 0.01), with positive contrast values, indicating a continuous growing utilization rate. UMC2 demonstrated the same pattern as the intervention hospital: no significant differences within the pre- and the post-measurements, but then again a significant and positive result for the ‘intervention’ (P < 0.01).

Figure 2 shows the percentage of operating rooms running after scheduled opening hours. As predicted by the simulation model, this percentage increased after the intervention in the Erasmus MC. This was confirmed by the contrast analyses where no significant change between the periods before the intervention and was followed by a significant change between pre- and post-measurements (P < 0.01). However, to attribute the change uniquely to the intervention, a non-significant difference between the two post-measurements was expected, but this was not confirmed since the difference was significant (P < 0.04). Moreover, for both control UMCs, the same pattern occurred: steady pre-measurements and significant increase in the percentage of OR’s running late after the intervention and between the two post-measurement periods. Therefore, the increase of the percentage of OR’s running after scheduled time was confirmed, but the question remains whether this could be attributed uniquely to the intervention.

The third and last relevant performance indicator was cumulative ‘overtime’. Figure 3 shows an increase in overtime in the Erasmus MC during the study period. From the contrast analysis (Table 2), all contrast analyses were significant (P < 0.05). This means a continuing change in overtime during the study period. Figure 3 and Table 1 show that the direction of the change was contrary to the expectations of Wullink et al.12. The contrast analysis gave significant results for all comparisons.

In addition, in the Erasmus MC a mean of 33.74% of all cancellations per year were cancellations to accommodate an emergency case, compared to a mean of 20.96% and 10.33% in the two control centers.

Table 1. Measurements of performance indicators (mean per year, per UMC)

Raw utilization (%) Number of ORs running after scheduled room exit time (% of total staffed ORs)

Cumulative ‘overtime’ (% of total allocated block time)

EMC UMC1 UMC2 EMC UMC1 UMC2 EMC UMC1 UMC2

2000-2001 66.87% 68.75% 64.04% 37.38% 49.69% 31.29% 5.95% 8.87% 6.32%

2002-2003 65.70% 71.37% 66.13% 38.39% 47.91% 29.05% 6.48% 7.72% 5.92%

2004-2005 67.68% 77.06% 70.45% 39.08% 52.09% 31.33% 7.33% 7.88% 5.72%

2006-2007 68.78% 80.67% 69.19% 41.17% 51.38% 31.78% 7.81% 7.55% 5.55%

193

Ope

ratin

g R

oom

Cap

acity

for E

mer

genc

y S

urge

ry

Table 2. Results of ANOVA-contrast analysis

Raw utilization (%) Number of ORs running after scheduled room exit time (% of total staffed ORs)

Cumulative ‘overtime’ (% of total allocated block time)

value p-value value p-value value p-value

EMC pre-measurements -0.01 0.33 0.01 0.31 0.01 0.03

post-measurements 0.01 0.29 0.02 0.04 0.01 0.05

intervention 0.04 0.01 0.04 0.01 0.03 0.01

UMC1 pre-measurements 0.03 0.02 -0.02 0.06 -0.01 0.01

post-measurements 0.04 0.01 -0.01 0.51 0.01 0.11

intervention 0.18 0.01 0.06 0.01 -0.01 0.01

UMC2 pre-measurements 0.02 0.08 -0.02 0.01 0.01 0.04

post-measurements -0.01 0.29 0.01 0.60 0.01 0.41

intervention 0.09 0.01 0.03 0.02 -0.01 0.01

Figure 1. Raw utilization (%) of the OR departments, mean per year per UMC

55%

60%

65%

70%

75%

80%

85%

90%

95%

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Raw utilization (%)

EMC UMC1 UMC2

194

10

Figure 2. Number of ORs running after scheduled room exit time (% of total staffed ORs), mean per year per UMC

Figure 3. Cumulative ‘overtime’ (% of total allocated block time), mean per year per UMC

25%

30%

35%

40%

45%

50%

55%

60%

65%

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

OR´s running after scheduled room exit time (%)

EMC UMC1 UMC2

5%

6%

7%

8%

9%

10%

11%

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Cumulative 'overtime' (%)

EMC UMC1 UMC2

195

Ope

ratin

g R

oom

Cap

acity

for E

mer

genc

y S

urge

ry

DISCUSSION

This study shows that in daily practice a dedicated OR for emergency cases is preferred over the approach of evenly reserving capacity for emergency surgery in all elective ORs, in performance terms regarding raw utilization (%), ‘overtime’ and the number of ORs running late. Moreover, the additional data concerning case cancellations indicates that a dedicated emergency OR has the benefit of less case cancellations.

The controlled time-series design/ANOVA with contrast analyses using empirical OR data (2000 - 2007) showed that after closing the dedicated emergency OR in the Erasmus MC, the number of ORs with overtime, as well as total overtime significantly increased. It is interesting to note that these results are partially in contrast with the simulation study by Wullink et al.12, which concluded that total overtime would decrease after closing the emergency OR. It was also simulated that OR utilization would increase slightly more (+1%) compared to what the empirical data showed. The results associated with the relative increase in the number of ORs with overtime correspond with the simulated results.

Remarkably, the analyses that were performed using empirical OR data of two control UMCs while maintaining a dedicated emergency OR, showed an even higher significant increase in OR utilization. No significant change in the number of ORs with overtime was found. However, a significant decrease in total overtime was revealed in both control UMCs during the same period (2000 - 2007), that was labeled as the ‘intervention contrast’ (intervention in Erasmus MC).

This study assessed the change in efficiency parameters of the problem that was simulated by Wullink et al.12. The results of their specific simulation study led to closing of the emergency OR in the Erasmus MC and evenly allocating emergency capacity in all elective ORs. However, the results of this recent study are partially in contrast with the simulated results: in retrospect, overtime significantly increased after the intervention. These recent results were based on the empirical data. OR utilization did increase, nevertheless, this increase was lower than the increase in utilization found in both control UMCs without the specific intervention of closing the emergency ORs. Therefore, these recent results do not support the earlier published conclusions5, 6, 11, 12 that distributing free OR capacity for emergency surgery evenly over all elective ORs performs better than dedicated emergency ORs on measures regarding the efficient use of scarce OR time.

Fone et al.14 already concluded that the potential of simulation modeling to inform evidence-based policy is clear, however, information on the outcomes of model implementation and henceforward the value of modeling requires further research. The earlier published conclusions5, 6, 11, 12 which led to closing the emergency OR in the Erasmus MC, were decently based on simulated results. One major drawback of this approach is that simulation results do not always reflect the discouraging reality, which was the case in this recent study. Earlier

196

10

findings in favor of a dedicated emergency OR were based on either simulation modeling8, 10, or empirical data analyzed with a straightforward “before-after” approach3, 4, 9, 13. “Before-after” designs suffer two disadvantages: first, background factors can produce large fluctuations in processes or outcomes of interest unrelatedly to the specific intervention that is studied; second, in time, multiple changes occur within a health care system or its socioeconomic environment and one or more of these other changes might have produced the preferred improvements16.

If, in this recent study, merely a “before-after” design was applied, two out of the three performance variables were confirmed, corresponding to the simulated results. Nevertheless, by strengthening the research design with two control UMCs, these effects turned out to be debatable. There are two ways to minimize the drawbacks of simple “before-after” designs, i.e. a time-series design with multiple time periods and a controlled before-after design, in which the same measurements occur in one or more hospitals that did not implement the intervention of interest. These two designs convey the extent of background variation. In the current study, both designs were combined, which resulted in applying a quasi-experimental controlled time-series design (ANOVA with four contrasts, as well as two control UMCs).

Nevertheless, this study as well as its design, has a number of limitations. First, there is a precision deficiency with regard to the scheduled OR start and finish times for the data collected in the years before 2004. In 2004, the OR departments of all eight UMCs in the Netherlands established a benchmarking collaborative18-21. Since then, each UMC provides surgical case records extracted directly from the hospital’s self-reported OR data management system to a central OR benchmark database. In this collaborative during the first two years, data definitions of OR time periods, uniform methods of data registration, and definitions of performance indicators were developed and harmonized among all eight participating UMCs. During this harmonization phase the adjustment of the common scheduled OR starting and finishing time was ensured in all UMCs, in case of an intentionally extended starting or finishing time due to e.g. team meetings or extended allocated block time for long procedures (taking over 4 hours per procedure). Calculations before 2004 were based on the common scheduled times. This limitation, however, is not considered to be of great impact on the results because deviating from common scheduled times is rather an exception.

Even though a controlled time-series design was applied, a second limitation is that this study with empirical data was not able to exclude all background variation. OR departments usually are departments subject to several projects, organizational developments and (technical) innovations at once, due to the multidisciplinary nature of the department itself and the ‘clients’ (all different surgical departments using the OR facilities). Possible background variations could be the fluctuation of the number of OR sessions (one session is one patient/case), the mean case duration, and the percentage of emergency patients, which periodically differs per year in one UMC or differs between UMCs. A third and last limitation is that data

197

Ope

ratin

g R

oom

Cap

acity

for E

mer

genc

y S

urge

ry

were gathered in tertiary referral centers only, and therefore, general applicability of the findings may be restricted to this specific academic segment. The ratio emergency versus elective patients as well as the complexity of emergency patients could be dissimilar from non-academic facilities.

As a general rule, there are boundaries to how far computational experiments are able to reflect the complexities of organizations and the people working there. In other words, the theory is right but practice is different. Simulation models as referred to earlier11, 12 described that they were not able to consider all practical constraints, hence, the model was an ‘abbreviated’ version of reality. Wullink etal.12 specifically mention that “implementation of the policy by which emergency capacity is reserved in all elective ORs, requires all stakeholders on the OR to strictly adhere to the policy”. The OR managers in the Erasmus MC corroborated that, in practice, not all stakeholders participated and not all surgical departments reserved free OR capacity for emergency surgery in their elective program as agreed upon. This hampers the performance of the approach that evenly allocates capacity for emergency surgery to all elective ORs considerably, as this model is most dependent on the collaboration and commitment of all surgical departments.

Future research on the outcomes of model implementation in health care settings such as the multidimensional OR environment, is required to contribute to closing the gap between operations research and organizational practices14, 22. Similar to Evidence Based Medicine, barriers and bridges exist to completely implement Evidence Based Management as a quality movement with the goal to explicitly apply the best and contemporary evidence in management and decision-making22-24. A critical barrier is the lack of time in daily practice to search, read, analyze, interpret and implement available evidence. Moreover, in contrast with clinicians, managers are not educated to know or use scientific evidence22.

A reasonable approach for future interventions in the multifaceted OR environment could be to follow-up policy decisions that are based on simulation modeling results with a robust controlled time-series research design, in which the change in parameters is carefully assessed. In line with the well-known Deming’s cycle, every implementation should be followed-up by an evaluation. To promote Evidence Based Management researchers need to provide more information regarding the specific context in which an intervention lead to a particular specific outcome22, 23. More detailed description of assumptions used in a simulation model might help bridging the gap between theory and practice14. Everything considered, the findings of this study do not support strong recommendations to close dedicated emergency ORs, as was recommended earlier based on simulated results, since information on the outcomes of model implementation is missing.

198

10

R EFERENCES



3. Heng M, Wright JG. Dedicated operating room for emergency surgery improves access and efficiency. Can J Surg 2013;56:167-74.

4. Bhattacharyya T, Vrahas MS, Morrison SM, et al. The value of the dedicated orthopaedic trauma operating room. J Trauma 2006;60:1336-40; discussion 40-1.

5. Bowers J, Mould G. Managing uncertainty in orthopaedic trauma theatres. European Journal of Operational Research 2004;154:599-608.

6. Brasel KJ, Akason J, Weigelt JA. Dedicated operating room for trauma: a costly recommendation. J Trauma 1998;44:832-6; discussion 6-8.


8. Ferrand Y, Magazine M, Rao U. Comparing two operating-room-allocation policies for elective and emergency surgeries. Proceedings of the 2010 Winter Simulation Conference 2010:2364-74.

9. Lovett BE, Katchburian MV. Emergency surgery: half a day does make a difference. Ann R Coll Surg Engl 1999;81:62-4.

10. Persson MJ, Persson JA. Analysing management policies for operating room planning using simulation. Health Care Manag Sci 2010;13:182-91.

11. Van Essen JT, Hans EW, Hurink JL, Overberg A. Minimizing the waiting time for emergency surgery. Operations Research for Health Care 2012;1:34-44.

12. Wullink G, Van Houdenhoven M, Hans EW, van Oostrum JM, van der Lans M, Kazemier G. Closing emergency operating rooms improves efficiency. J Med Syst 2007;31:543-6.

13. Lemos D, Nilssen E, Khatiwada B, et al. Dedicated orthopedic trauma theatres: effect on morbidity and mortality in a single trauma centre. Canadian Journal of Surgery 2009;52:87-91.


15. Fitzgerald J, Lum M, Dadich A. Scheduling unplanned surgery: a tool for improving dialogue about queue position on emergency theatre lists. Aust Health Rev 2006;30:219-31.




19. van Veen-Berkx E, Bitter J, Elkhuizen SG, et al. The influence of anesthesia-controlled time on operating room scheduling in Dutch university medical centres L’influence du temps controle par l’anesthesie sur le programme operatoire dans les centres medicaux d’une universite neerlandaise. Can J Anaesth 2014;61:524-32.

199

Ope

ratin

g R

oom

Cap

acity

for E

mer

genc

y S

urge

ry

20. van Veen-Berkx E, Bitter J, Elkhuizen SG, et al. Comment on Research Article Entitled “Variability of Subspecialty-Specific Anesthesia-Controlled Times at Two Academic Institutions” as published in J Med Syst 2014; 38 (11). J Med Syst 2014;38:51.


22. Rousseau DM. Is There Such a Thing as “Evidence-Based Management”? Academy of Management Review 2006;31:256-69.

23. Pfeffer J, Sutton RI. Evidence-Based Management. Harv Bus Rev 2006:1-13.

24. Sackett DL, Rosenberg WM, Gray JA, Haynes RB, Richardson WS. Evidence based medicine: what it is and what it isn’t. BMJ 1996;312:71-2.

Reflections

11Twelve Years of Collaborative Benchmarking

the Operating Room Departments of

Eight Dutch University Medical Centers

204

11

The joint initiative of the eight Dutch University Medical Centers (UMCs) “OR Benchmarking” was originated with the goal to learn, share knowledge and improve the performance of the OR departments. The results after twelve years of benchmarking can be divided into “tangible and intangible” (or “soft”) results. The series of performance indicators based on harmonized data definitions of OR time periods, uniform methods of data registration and harmonized definitions between the eight participating centers, in other words the “OR benchmarking measurement system”, is a key tangible result.

Other material results are the central OR Benchmark database in which longitudinal data of the eight centers is collected, along with its management reporting tool. The ongoing data collection started in 2005. Today (reference date 1 January 2015), the database comprises 365,151 OR-days on which a total of 1,374,363 surgical cases were performed. All benchmarking participants can access the OR benchmark data on any occasion, using a highly secured web-based management reporting tool.

The following graphs provide a general overview of the database.

Absolute number of surgical cases performedFigures 1a-c show the total number of surgical cases performed per center per year. This includes elective as well as non-elective (emergency) cases, inpatient cases, outpatient (ambulatory, same-day surgery) cases, as well as cases performed on sub-locations with their own separate OR department, such as Children’s Hospitals, a Cancer Center and an Eye Center. In one UMC it was not possible to include the data of the Thoracic Centre with a separate OR location, due to the use of an incompatible registration system. During two, respectively three, years, two UMCs were not able to provide data to the benchmarking collaborative, due to the transition to another Hospital Information System. On an overall level the eight UMCs together perform circa 137,436 surgical cases per year (a mean of 17,180 cases per center every year).

205

Twel

ve Y

ears

of O

pera

ting

Roo

m B

ench

mar

king

in th

e N

ethe

rland

s

Figure 1a1. Total number of surgical cases performed per UMC per year, including elective as well as non-elective cases, inpatient and outpatient cases, as well as cases performed on sub-locations (including Children’s Hospitals, Cancer Center, and Eye Center, excluding Thoracic Center). Data including all different surgical specialties in the UMCs.

Figure 1b. Total number of surgical cases performed per UMC per year, including elective as well as non-elective cases, including merely inpatient cases performed on the main OR department of each center. Data including all different surgical specialties in the UMCs.

1 UMC1 excluding data from year 2010; UMC6 excluding data from years 2011-2012.

206

11

Figure 1c. Total number of surgical cases performed per UMC per year, including merely elective, inpatient cases performed on the main OR department of each center. Data including all different surgical specialties in the UMCs.

Elective and non-elective surgical casesFigure 2 demonstrates the ratio of inpatient elective and non-elective cases per UMC. Each year this ratio is fairly constant. The ratio does differ per UMC: rates of elective cases ranging from 69%-83% and non-elective cases from 17%-31%. On an overall level of all eight UMCs the ratio is 75% (elective cases) as opposed to 25% (non-elective cases). The ratio may reveal information on the effectiveness of the short-term scheduling method regarding non-elective or emergency surgery and the access to surgery (waiting time). Emergency surgery is prioritized on the basis of clinical priorities1. Each priority is associated with a maximum period of time within surgical treatment should be provided. In the Netherlands, priorities A – E (or in other centers characterized as 1 – 5) are generally defined as:

A. Immediately life threatening (immediate surgery within 15 minutes);B. Organ or limb threatening (surgery within 2 hours);C. Delay in treatment likely to affect clinical outcome (within 6 hours);D. Delay in treatment increases risk for infections and length-of-stay in hospital (within 24

hours);E. Semi-urgent, patient not stable for discharge (within 72 hours).

Because of these different categories of clinical priorities, the OR Benchmark database merely makes a distinction between ‘elective’ and ‘non-elective’ surgical cases.

207

Twel

ve Y

ears

of O

pera

ting

Roo

m B

ench

mar

king

in th

e N

ethe

rland

s

Figure 2. Ratio of inpatient, elective versus inpatient, non-elective cases per UMC. Data year 2014, including all different surgical specialties in the UMCs. On an overall level of all eight UMCs the ratio is 75% (elective case) as opposed to 25% (non-elective cases).

The ratio also reflects the hospital’s catchment area and the type of patient case mix, since the amount of emergency cases affects OR scheduling and therefore the utilization of OR capacity. In general, two approaches prevail for reserving capacity for emergency surgery: (1) dedicated emergency ORs and (2) evenly allocating capacity to all elective ORs. Merely one UMC adopted the second approach by the end of 2007 as a result of a simulation study2 (see Chapter 10). Nowadays, this hospital adopted a semi-reversion to the first approach for surgical specialties performing a large number of emergency patients, such as the General Surgery department. And recently, another UMC switched from the first to the second approach. This is an important intervention that should be followed-up in future studies.

Duration of surgical casesFigure 3 demonstrates the mean duration of total procedure time in minutes per UMC per year, including purely inpatient (elective as well as emergency) surgical cases, representing the major part of the database. Total procedure time is the sum of anesthesia-induction time, surgeon-controlled time and anesthesia-emergence time3. From the chart, it can be seen that in practically all centers the mean duration of surgical procedures increased, which reflects the concentration of complex, highly specialized care within the UMCs.

208

11

Figure 3. Median duration of total procedure time in minutes per UMC per year, including all inpatient (elective and emergency) surgical cases. Data including all different surgical specialties in the UMCs.

Overall, every center demonstrated a structural underestimation of total procedure time, which is also demonstrated by Figure 4a and 4b. This further supports the idea of earlier research that surgeons tend to underestimate the time needed to finish a procedure4-7 and that anesthesiologists are not always capable to accurately predict the time needed for anesthesia8, 9. Estimation accuracy varies comprehensively and scheduling methods relying on surgeon- and procedure-specific historical case time averages are forces to use a very small sample. The typical university hospital OR is characterized by a rich variation in patient case mix, making it challenging to forecast total procedure time. Due to wide dispersion, use the historical median to generate the first estimate, subsequently, the surgeon can adjust that estimate based on patient complexity factors and OR team characteristics (e.g. surgical/anesthesia residents).

Our research as described in Chapters 3, 4, 6, 7, 8 and 9 confirm the improvement potential with regard to OR scheduling and minimizing prediction errors3, 10-12. OR scheduling is complex because a procedure entails several elements subject to variability, such as patient and OR-team characteristics, room setup and takedown, patient positioning, prepping and draping, as well as the two principal components: surgeon-controlled time and anesthesia-controlled time. Scheduling these two principal components of a procedure more accurately, leads to less case cancellations and lower prediction errors. This may result in a more efficient use of limited and expensive OR resources.

209

Twel

ve Y

ears

of O

pera

ting

Roo

m B

ench

mar

king

in th

e N

ethe

rland

s

Figure 4a. Stacked histogram of the prediction errors (actual – predicted total procedure time) in minutes, showed on the scale from -300 to +300 minutes. Data year 2014, including all different surgical specialties in the UMCs.

Figure 4b. Stacked histogram of the prediction error % ((actual – predicted total procedure time) / predicted time * 100%). Positive % meaning ‘underestimation’; the case took longer than predicted. Negative % meaning ‘overestimation’; the case took shorter than predicted. Data year 2014, including all different surgical specialties in the UMCs.

prediction error (in minutes)actual - predicted

300250200150100500-50-100-150-200-250-300

Fre

qu

en

cy

125.000

100.000

75.000

50.000

25.000

0

UnderestimationOverestimation

Page 1

prediction error (%)

200

190

180

170

160

150

140

130

120

110

100

90

80

70

60

50

40

30

20

10

0-10

-20

-30

-40

-50

-60

-70

-80

-90

-100

Fre

qu

en

cy

3.000

2.000

1.000

0

UnderestimationOverestimation

Page 1

210

11

Utilization of allocated operating room timeThe overall performance of one OR-day, generally equal to eight hours of block time (8:00h – 16:00h) allocated to a specific surgical department, is commonly evaluated by the indicator ‘raw utilization’13-15. It is a measure for the use of staffed operating room time and is defined as the total amount of time patients are present in the OR divided by the total amount of allocated block time per day × 100%. This definition excludes turnover time and overtime at the end of the day. In the central OR Benchmark database utilization is also measured including turnover time, however, we prefer to measure and evaluate turnover time as a separate indicator because, especially in a university hospital setting with large OR facilities and longer patient transport times, this can identify further areas for improvement.

The difference in raw utilization (median) between the ‘best and worst performer’, so to speak, of the eight UMCs is 8% in 2014, see also Figure 5a and 5b. Improving OR scheduling may increase the utilization of OR time with a maximum of 8%. Since differences between UMCs regarding raw utilization of OR time are not remarkable, there still is a 8% of improvement potential. However, a higher utilization percentage is not always preferable. In many cases, a high utilization rate will go hand in hand with overtime at the end of the day. Overtime may result in staff overtime payments, employee dissatisfaction as well as patient dissatisfaction in case of cancellations16.

Moreover, principal goals of effective operating room management are not merely efficiency but first of all safety and quality for patients and staff. This OR Benchmark database purely measures one, although essential, aspect of OR efficiency, i.e. the (non-) utilization of expensive and scarce OR time. Since the start in 2004, the collaborative organized two-monthly multidisciplinary focus-group study meetings (or networking events) to discuss the results of data analyses and explore the practices ‘behind the data’. Later on these meetings were also organized to discuss current topics high on the OR management agenda regarding patient safety and quality of care. For instance, the introduction and implementation of the WHO surgical safety checklist.

211

Twel

ve Y

ears

of O

pera

ting

Roo

m B

ench

mar

king

in th

e N

ethe

rland

s

Figure 5a. Raw utilization (%) per UMC per year, including all inpatient OR-days. Note that y-axis starts at 60%. Data up to and including second quarter (Q2) of 2015. Data including all different surgical specialties in the UMCs.

Figure 5b. Boxplot of raw utilization (%) per UMC in 2014, including all inpatient OR-days. Data including all different surgical specialties in the UMCs. Early start (entry time of the first patient on that day before the scheduled start of allocated block time, generally 8:00h AM) and processing patients in a parallel fashion, can result in a raw utilization percentage of more than 100%.

212

11

Returning at the OR Benchmark database: differences between UMCs regarding raw utilization of OR time are not remarkable, nevertheless, interorganizational learning opportunities do exist. These opportunities derive from performance indicators measuring non-utilization of OR time, as described in Chapter 2 and below in addition to the overview of the database.

First-case tardinessIt is common for OR management to spend effort on reducing first-case tardiness because of the expected, positive, psychological effect on the OR team and on the workplace of ‘starting on time’ throughout the whole day. Starting on time means less rush, which is one of the conditions leading to a safe working environment14. Moreover, OR management considers an alleged ‘trickle down’ effect that a late start of the first case of the day causes all subsequent cases to start late14, 15. In our study with regard to ‘non-operative time’ (late start, turnover time and underused time at the end of the day), however, this specific effect was not supported.

Figure 6. Boxplot of first-case tardiness in minutes per UMC, data year 2005 compared to year 2015, including all inpatient OR-days, data 2015 including first (Q1) and second quarter (Q2). Data including all different surgical specialties in the UMCs.

The OR benchmark data of the eight university hospitals represented in Figure 6, reveal that in almost all centers the median or maximum duration of first-case tardiness decreased, comparing the start of the benchmarking collaborative in 2005 with the most recent data

213

Twel

ve Y

ears

of O

pera

ting

Roo

m B

ench

mar

king

in th

e N

ethe

rland

s

collection in 2015 (Q1 and Q2). The variation in first-case tardiness (duration in minutes) reduced from an interquartile range (IQR) of 23 minutes in 2005 to an IQR of 15 minutes in 2015. Variation is undesirable: it creates uncertainty in the ability to generate a desired outcome. This originates from the Lean Six Sigma philosophy17. Six Sigma focusses on reducing process variation and enhancing process control, where Lean drives out waste (non-value-added) and stimulates work standardization and process flow. A decrease of the IQR in first-case tardiness during the years indicates a consistent and stable process, as well as an organizational learning effect.

All performance indicators measured in this benchmark are characterized by a positively-skewed lognormal distribution. Therefore, data is represented in ‘boxplots’18: at the midpoint of the plot is the median (a bold black line), which is surrounded by a box, the top and bottom of which are the limits within the middle 50% of observations are located. The interquartile range is also called the “middle 50” and is a measure of dispersion. It is calculated by subtracting the upper and lower quartiles: IQR = Q3 - Q1. The whiskers extend from the first quartile to the minimum value and from the third quartile to the maximum value. A more compact boxplot indicates the achievement of less process variation.

Combining the duration in minutes with the mean frequency of occurrence (Table 1), one could argue that the so-called ‘academic fifteen minutes’ still endures in the OR departments of the Dutch UMCs. We also refer to our study in Chapter 5, which describes that four centers implemented successful interventions to reduce first-case tardiness14.

Table 1. Mean frequency of occurrence per OR-day, data year 2005 compared to year 2015 (Q1 and Q2), including all inpatient OR-days. Data including all different surgical specialties in the UMCs. This frequency has to be interpreted as the percentage of all staffed operating rooms on a random weekday starting too late. Example UMC1: 88% of all staffed ORs on a random weekday start 13 minutes (median) too late.

Mean frequency first-case tardiness (“late start”)

2005 2015

UMC1 0.70 0.88

UMC2 0.63 0.71

UMC3 0.38 0.63

UMC4 0.79 0.55

UMC5 0.50 0.62

UMC6 0.57 0.59

UMC7 0.69 0.68

UMC8 0.85 0.66

214

11

Turnover timeOn an overall level of all UMCs, the performance in terms of turnover time per case/per patient slightly increased from 13 to 16 minutes (median). An essential part of turnover time is room cleaning time, transportation time as well as set-up time for the next procedure, and therefore this indicator can never be reduced to zero. However, reducing delay time and possible transportation time could result in shorter turnover time, which is important because the longer the turnover time, the lesser the raw utilization of the available OR time15 (Chapter 2).

Figure 7. Boxplot of turnover time per case in minutes per UMC, data year 2005 compared to year 2015, including all inpatient OR-days, data 2015 including first (Q1) and second quarter (Q2). Data including all different surgical specialties in the UMCs.

Table 2. Mean turnover frequency per OR-day, data year 2005 compared to year 2015 (Q1 and Q2), including all inpatient OR-days. Data including all different surgical specialties in the UMCs. Turnover frequency has to be interpreted as the mean number of cases performed in every staffed operating room per random weekday.

Mean turnover frequency

2005 2015

UMC1 1.78 1.34UMC2 1.54 1.46UMC3 1.38 1.28UMC4 1.75 1.73UMC5 1.63 1.54UMC6 1.17 1.28UMC7 2.41 3.85UMC8 1.58 1.88

215

Twel

ve Y

ears

of O

pera

ting

Roo

m B

ench

mar

king

in th

e N

ethe

rland

s

The duration of turnover time in minutes is related to the turnover frequency per operating room per day (Table 2) and thus related to the duration of total procedure time. The shorter the duration of procedures in one OR, the higher the frequency of turnovers and the higher the cumulative turnover time in minutes per OR-day. UMC7 demonstrates the highest turnover frequency combined with the shortest duration of total procedure time per case (Figure 3) and the highest absolute number of elective, inpatient cases performed (Figure 1c).

Under- and overutilized timeFigure 8 shows that in almost all UMCs empty OR time (or underutilized OR time) at the end of the day marginally decreased: from a median duration of 51 minutes (IQR 75) in 2005 to 43 minutes (IQR 69) in 2015. On an overall level of all centers, overtime (or overutilized time) at the end of the day showed a practically similar level of performance in 2005 and 2015: a median duration of 40 minutes and IQR of 61 minutes (Figure 9). In contrast with the start of an OR-day, just a small amount of surgical cases will end exactly on the scheduled room exit time being the scheduled end of allocated block time, thus, almost every last case of the day will result in either empty OR time or overtime. The duration of empty OR time and overtime should be as low as possible and the frequency of occurrence about 50% for both as a result of an OR schedule ‘in balance’ (Table 3).

Figure 8. Boxplot of empty OR time in minutes per UMC, data year 2005 compared to year 2015, including all inpatient OR-days, data 2015 including first (Q1) and second quarter (Q2). Data including all different surgical specialties in the UMCs.

216

11

Figure 9. Boxplot of overtime in minutes per UMC, data year 2005 compared to year 2015, including all inpatient OR-days, data 2015 including first (Q1) and second quarter (Q2). Data including all different surgical specialties in the UMCs.

Table 3. Mean frequency of occurrence of empty OR time and overtime per OR-day, data year 2005 compared to year 2015 (Q1 and Q2), including all inpatient OR-days. This frequency has to be interpreted as the percentage of all staffed operating rooms with either empty OR time or overtime at the end of a random weekday.

Mean frequency empty OR time Mean frequency overtime

2005 2015 2005 2015

UMC1 0.50 0.55 0.46 0.35

UMC2 0.57 0.53 0.37 0.39

UMC3 0.48 0.68 0.45 0.22

UMC4 0.42 0.40 0.51 0.55

UMC5 0.49 0.45 0.44 0.44

UMC6 0.55 0.50 0.34 0.42

UMC7 0.53 0.55 0.42 0.40

UMC8 0.61 0.56 0.33 0.35

In the study with regard to ‘non-operative time’15 (Chapter 2) we demonstrated that empty OR time at the end of the day had the strongest influence on raw utilization. A long duration of empty OR time as well as a high frequency of occurrence means that expensive OR time is poorly utilized and possibly more surgical cases could have been performed during that idle time.

217

Twel

ve Y

ears

of O

pera

ting

Roo

m B

ench

mar

king

in th

e N

ethe

rland

s

On the other hand, a high frequency and long duration of overtime will be profitable with regard to OR utilization, yet less profitable for OR personnel and overtime costs. By offering fresh OR staff for each shift, overtime becomes less of a disadvantage. UMCs as well as different OR locations – inpatient and outpatient – can intentionally contrast in their policy concerning the balance between underutilized and overutilized time at the end of the day, which is reflected in the performance indicators. Nevertheless, measuring the performance at the end of the day can evaluate the results of chosen policies and benchmarking creates learning opportunities on this particular subject.

The performance of the end of an OR-dayIn general, three indicators evaluate the performance of ‘the end of an OR-day’. The end of one day balances between either empty OR time or overtime, along with the potential cancellations of elective surgical cases15. Unfortunately, the cancellation rate was not structurally included as one of the performance indicators in the OR benchmark database because it was not possible to overcome crucial differences in registration methods and definitions of cancellations and reasons. Cancellations interrupt patient flow and result in loss of revenue for hospitals, which is why the cancellation rate is a common metric in OR dashboards for measuring performance. Therefore, cancellations were benchmarked a couple of times ‘ad hoc’ to create learning opportunities between the UMCs.

A specific analysis was executed concerning the connection between these three indicators because of their presumable conflicting interests; the concept of the ‘devil’s triangle’ – a familiar concept in project management – could also apply in the OR department. According to this triangle the three constraints (empty OR time, overtime and cancellations) in the corners are mutual dependent and when one of the three constraints changes, the other two also have to change. Undoubtedly the OR Benchmark data showed a relationship between empty OR time and overtime at the end of the day. Additional analyses demonstrated more empty OR time when the cancellation rate increased and overtime decreased. Overtime increased when cancellations were low. Again, these metrics and relations evaluate the effects of policymaking, e.g.:

• a ‘zero tolerance for overtime’ policy in favor of supporting OR-staff and minimizing overtime costs;

• a ‘zero tolerance for case cancellations’ policy in favor of patient satisfaction; • a ‘zero tolerance for empty OR time’ policy in favor of surgeon satisfaction and

utilization of OR time.

Surgical cases at nightA final indicator that was measured since the start of the OR benchmarking collaborative is ‘surgical cases performed at night’, including non-elective or emergency cases along with

218

11

elective cases performed between 00:00h and 07:00h. This indicator was first initiated in 2007 during the campaign called ‘Faster, Better’, a program of the Dutch Ministry of Public Health, Welfare and Sport, the Dutch Association of Hospitals and the Order of Medical Specialists, with the aim to improve transparency, efficiency and quality of healthcare. It is internationally preferred that non-essential night-time operating is reduced to a minimum with respect to patient safety and quality concerns. Previous research found that the overall postoperative mortality and complications were significantly higher during nights – as well as during holidays and weekends – than surgeries that were performed during daytime on weekdays19-21. Remarkably, this significant effect merely counted for elective cases and not for emergency cases20, 22, 23. Complementary research studied the impact of sleep deprivation on surgeon’s performance during night shifts24, which concludes that this impact is complex and multidimensional. Apparently, surgeons do feel a certain amount of impact of sleep loss and their ‘circadian body clock’ is affected. Nevertheless, surgeons are able to compensate for these effects since the research did not find a negative relationship with patient safety during night shifts.

Figure 102. Absolute number of inpatient, elective cases performed at night (between 00:00h and 07:00h) per UMC per year. Data including all different surgical specialties in the UMCs.

2 UMC1 excluding data from year 2010; UMC6 excluding data from years 2011-2012.

219

Twel

ve Y

ears

of O

pera

ting

Roo

m B

ench

mar

king

in th

e N

ethe

rland

s

Since 2005 the number of elective, inpatient cases performed at night was significantly reduced in almost all UMCs (Figure 10). The high production of cases during the night shown by large UMCs can be explained by the broad hospital’s catchment areas, as well as the exceptional medical procedures (so-called “Article 2” procedures) like transplantations merely these UMCs perform. On an overall level of all centers a maximum of 90% and a minimum of 78% of all inpatient, elective as well as non-elective cases, were performed during office hours (or allocated block time, generally between 08:00h and 16:00h), 5-10% was performed during the evening, 1-3% during nighttime and 4-10% during the weekend (Figure 11).

Figure 11. Percentage of cases per starting time per UMC: during allocated block time (“office hours”), during evening hours (between end of block time and 00:00h), during nighttime (between 00:00h and 07:00h) and during the weekend (on Saturday or Sunday). Data per UMC per year. Data year 2014, including all different surgical specialties in the UMCs.

Knowledge sharing network and future perspectives on OR benchmarkingAll the above-described graphs and numbers provided a summary of the main features that are (still) collected in the central OR Benchmark database, four times a year. The organizational characteristics and performance indicators in this database are also available per type of OR sub-location (as shown in Figure 1a), and also per medical specialism (surgical department).

220

11

The research described in this thesis mainly focused on one organizational level, i.e.: the largest or main OR departments predominantly performing inpatient cases for all different surgical departments. In the Netherlands, the outpatient surgery workflow is usually performed in a separate sub-location of the OR department and is considered as a distinct process. Though for some years a shift from inpatient to outpatient surgery is noticed, it would be interesting to analyze the numbers and performance of outpatient surgery in a university hospital setting in the future, supported by this OR Benchmark database. In contrast to the rest of this thesis, the analyses in Chapters 8 and 9 focused on two specific surgical departments.

More research is required to determine improvement potential and learning opportunities with regard to benchmarking the OR performance of specific surgical specialties. The recent database does provide the relevant numbers per surgical specialty in every UMC, however, due to registration differences, it is not always possible to compare specific surgical procedures. In the future, it would be interesting to compare the OR performance of specific academic surgical procedures (e.g. organ transplantations) between the eight centers, supported by the existing OR Benchmark network.

Future work should not only explore OR performance on this ‘procedure level’ but also the value in terms of patient outcomes, safety and experiences, in order to connect with the actuality of current policies of Value-Based Health Care25. In this respect, our collaborative could learn from the examples set by the Dutch Institute for Clinical Auditing (DICA) including systematic audits and feedback of information about the process as well as outcomes to improve the quality of surgical care. Benchmarking outcomes on a national level for specific medical conditions, adjusted for case-mix differences, has identified best practices as well as negative outliers. The Association of Surgeons in the Netherlands initiated an independent audit committee to provide consultative advice to hospitals identified as negative outliers. This aspect of the Dutch clinical auditing system has contributed to substantial improvements in quality of care26. Other examples of measuring the outcomes (value) of perioperative care are the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP)27, the quality measurement tools of the Anesthesia Quality Institute in Illinois28, and the Perioperative Nursing Data Set of the Association of periOperative Registered Nurses (PNDS AORN)29.

Our qualitative study (Chapter 1) assessed the purposes, indicators, participating organizations and performance management system of OR Benchmarking, and found that collaborative benchmarking has positive, however intangible, side effects in addition to the actual performance comparison and benefits of performance improvement13. Important advantages are the business social network that was built and that benchmarking generates discussions about everyday challenges in operating room management and practice. Although the performance indicators mainly focus on the (non-) utilization of OR time, they were used to initiate and to facilitate discussions about a variety of subjects, such as quality of care,

221

Twel

ve Y

ears

of O

pera

ting

Roo

m B

ench

mar

king

in th

e N

ethe

rland

s

patient safety, workforce and management strategies. As described above, the main goals of effective operating room management are threefold: safety and quality for patients and staff, as well as efficiency. Therefore, the OR Benchmark collaborative repeatedly organized multidisciplinary focus-group study meetings (networking events) to discuss current topics high on the OR management agenda. These topics are listed in Table 4. The list is long but not comprehensive.

Table 4. Current topics examined and discussed in the Dutch OR Benchmarking collaborative in addition to the structural OR performance indicators regarding OR efficiency

Current topics examined and discussed in the Dutch OR Benchmarking collaborative

WHO Surgical Safety Checklist and Time-Out Procedure.

Surgical Site Infections (SSI) and the bundle of four interventions to reduce SSI’s: hygiene discipline, antibiotic prophylaxis, appropriate hair removal methods, perioperative normothermia3.

TeamSTEPPS teamwork and culture program in the OR.

“Game rules” in OR scheduling: communication and relationship between the OR and surgical teams.

Incident reporting and PRISMA-analyses in the OR.

Patient’s Satisfaction with perioperative care.

Management reporting OR and surgical teams (on strategic and tactical management levels).

Reasons for case cancellations on the day of intended surgery.

High-Risk Medication Safety in the OR (including the ‘double-check’ and parenteral medication)4.

Maintaining perioperative normothermia in surgical patients.

Optimization of sterile processing practices, tracking & tracing of instruments.

Procurement and Inventory Management in the OR.

Management Tools for OR Risk Management.

Human Resource Management, staffing/workforce expenses and staff productivity in the OR.

Demographics Analysis (Market Research) and discussion on concentration of low volume, high complex care.

Development and construction of new OR facilities.

Evidence-based Change Management in the OR.

Evidence-based OR Management.

3 This is one element of the VMS Dutch Patient Safety Program from the Dutch Ministry of Healthcare and the Health Care Inspectorate.

4 This is also one element of the VMS Dutch Patient Safety Program from the Dutch Ministry of Healthcare and the Health Care Inspectorate.

222

11

Study meetings were usually visited by 25 to 30 professionals per meeting and originating from all eight UMCs; these professionals represented OR management, anesthesiologists, surgeons, OR nurses, anesthesia nurses and staff advisors. These meetings helped to build a comprehensive and strong business-social network. Also nine nationwide one-day conferences, were organized. A total of 1,708 participants visited these conferences, originating from university medical centers as well as general and top-clinical hospitals.

The Dutch OR Benchmark collaborative still continues today and recently the steering committee decided (on the 23th of November 2015) to financially support the collaborative for yet another year. In 2016, the challenge to concretize the interorganizational learning opportunities deriving from our benchmarking activities, remains. The steering committee has stated that there will be two main approaches to take on that challenge: a) two or more centers will participate in a collaborative model for achieving breakthrough improvement on OR scheduling (based on the ‘Institute for Healthcare Improvement Breakthrough Series’ model, also described in Chapter 1); b) two or more centers will participate in a collaborative model to benchmark patient safety data regarding the SSI-bundle to create learning and improvement prospects on quality and safety issues in the OR. Both approaches will be dealt with either independently within the existing network or in cooperation with an external consultancy firm to drive more change and commitment.

223

Twel

ve Y

ears

of O

pera

ting

Roo

m B

ench

mar

king

in th

e N

ethe

rland

s

REFERENCES

1. NSW Department of Health SST. Emergency Surgery Guidelines; 2009.



4. Alvarez R, Bowry R, Carter M. Prediction of the time to complete a series of surgical cases to avoid cardiac operating room overutilization. Can J Anaesth 2010;57:973-9.


6. Dexter F, Wachtel RE, Epstein RH, McIntosh C, O’Neill L. Allocative efficiency vs technical efficiency in operating room management. Anaesthesia 2007;62:1290-1; author reply 1-2.



9. Travis E, Woodhouse S, Tan R, Patel S, Donovan J, Brogan K. Operating theatre time, where does it all go? A prospective observational study. BMJ 2014;349:g7182.

10. Bitter J, van Veen-Berkx E, Gooszen HG, van Amelsvoort P. Multidisciplinary teamwork is an important issue to healthcare professionals. Team Performance Management 2013;19:263-78.

11. Bitter J, van Veen-Berkx E, van Amelsvoort P, Gooszen H. Preoperative cross functional teams improve OR performance. J Health Organ Manag 2015;29:343-52.

12. van Veen-Berkx E, Bitter J, Kazemier G, Scheffer GJ, Gooszen HG. Multidisciplinary Teamwork Improves Use of the Operating Room: A Multicenter Study. J Am Coll Surg 2015.

13. van Veen-Berkx E, de Korne DF, Olivier OS, Bal RA, Kazemier G. Benchmarking Operating Room Departments in the Netherlands: Evaluation of a Benchmarking Collaborative between Eight University Medical Centres. Benchmarking: An International Journal 2016;23.




17. George ML. Lean Six Sigma: Combining Six Sigma Quality with Lean Production Speed: McGraw-Hill Education; 2002.


19. Phatak UR, Chan WM, Lew DF, et al. Is nighttime the right time? Risk of complications after laparoscopic cholecystectomy at night. J Am Coll Surg 2014;219:718-24.

224

11

20. Calland JF, Adams RB, Benjamin DK, Jr., et al. Thirty-day postoperative death rate at an academic medical center. Ann Surg 2002;235:690-6; discussion 6-8.

21. Wu JX, Nguyen AT, de Virgilio C, et al. Can it wait until morning? A comparison of nighttime versus daytime cholecystectomy for acute cholecystitis. Am J Surg 2014;208:911-8; discussion 7-8.

22. Turrentine FE, Wang H, Young JS, Calland JF. What is the safety of nonemergent operative procedures performed at night? A study of 10,426 operations at an academic tertiary care hospital using the American College of Surgeons national surgical quality program improvement database. J Trauma 2010;69:313-9.

23. van Zaane B, van Klei WA, Buhre WF, et al. Nonelective surgery at night and in-hospital mortality: Prospective observational data from the European Surgical Outcomes Study. Eur J Anaesthesiol 2015;32:477-85.

24. Amirian I. The impact of sleep deprivation on surgeons’ performance during night shifts. Dan Med J 2014;61:B4912.

25. Porter ME, Teisberg EO. Redefining competition in health care. Harv Bus Rev 2004;82:64-76, 136.

26. Van Leersum NJ, Snijders HS, Henneman D, et al. The Dutch surgical colorectal audit. Eur J Surg Oncol 2013;39:1063-70.

27. Hall BL, Hamilton BH, Richards K, Bilimoria KY, Cohen ME, Ko CY. Does surgical quality improve in the American College of Surgeons National Surgical Quality Improvement Program: an evaluation of all participating hospitals. Ann Surg 2009;250:363-76.

28. Whitlock EL, Feiner JR, Chen LL. Perioperative Mortality, 2010 to 2014: A Retrospective Cohort Study Using the National Anesthesia Clinical Outcomes Registry. Anesthesiology 2015.

29. Westendorf JJ. Utilizing the Perioperative Nursing Data Set in a surgical setting. Plast Surg Nurs 2007;27:181-4.

225

Twel

ve Y

ears

of O

pera

ting

Roo

m B

ench

mar

king

in th

e N

ethe

rland

s

12Summary and General Conclusions

General Limitations

General Discussion and Future Perspectives for Research

Lessons Learned

228

12

Summary and General Conclusions

The general aim of this thesis was to find an answer to the question whether a nationwide long-term benchmarking collaborative of the operating room (OR) departments of all eight University Medical Centers (UMCs) in the Netherlands could lead to improvements in overall OR management. For this purpose, several studies were conducted: one exploratory study combining qualitative and quantitative research methods; three descriptive studies based on a substantial amount of multicenter data; and six quasi-experimental studies to determine the effect of specific interventions in

different OR processes.

Benchmarking OR Departments in the NetherlandsTo investigate whether the collaborative and long-term approach of the Dutch OR benchmarking initiative has led to benefits in OR management, an evaluation frame based on literature1, 2 was applied in a mixed-methods study design (Chapter 1). Collaborative benchmarking has benefits different from mainly performance improvement and identification of performance gaps. It is interesting to note that, since 2004, the OR benchmarking initiative still endures after already existing for more than ten years. A key benefit was pointed out by all respondents as ‘the purpose of networking’. The networking events organized by the collaborative were found to make it easier for participants to contact and also visit one another in the OR departments of the eight university hospitals. Apparently, such informal contacts are helpful in spreading knowledge, sharing policy documents and initiating improvements in overall OR management. One reason for this is that they could be used to discuss the tacit components of best practices, that are hard to share in more formal communication media. Respondents were satisfied with the content of these meetings and with the exchange of knowledge in an informal manner, the exchange of experiences including sharing best practices as well as discussing worries and today’s challenges in OR management. It enables understanding and learning from each other. These findings corroborate the idea of De Korne etal. 1, 2 that participating in benchmarking offers other advantages, such as generating discussions about how to deliver services and increasing the interaction between participants.

During the initiation phase of the benchmark collaborative, a considerable amount of time (two years) and effort was undertaken by the steering committee to develop a collaboration agreement. This agreement created the foundation for trust and confidentiality between the eight participating partners, because confidentiality and ownership of benchmarking data are two delicate and important parts of the agreement. These first years were also seized by the development and harmonization of definitions of performance indicators. Common definitions are an essential base for external benchmarking 3, 4. The long-term commitment of the eight centers to the OR benchmark collaborative is exceptional, yet might also be necessary to build

229

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

and maintain trust between the centers, and also be necessary for uniform data registration and harmonization of indicator definitions.

Benchmarking is defined as a ‘continuous process’ 5 and encourages the use of a continuous quality improvement model (the PDSA cycle). Although this OR benchmark initiative, as many benchmark initiatives 6, started with a stated aim to improve, actual (measurable) quality or performance improvements are not necessary for this initiative to endure. These findings further support the idea of De Korne etal. 1, 2 that benchmarking is relying on iterative and social processes in combination with structured and rational process of performance comparison. The relatively limited focus on OR utilization in this benchmark seems to be a starting point for exchanging a variety of information and experiences considering the structure, process and performance of OR departments. More attention needs to be given to the relation between benchmarking as instrument and the actual performance improvements realized through benchmarking in the local UMC’s. A collaborative approach in benchmarking can be effective because participants use its knowledge-sharing infrastructure which enables operational, tactical and strategic learning. Organizational learning is to the advantage of overall operating room management. Benchmarking seems a useful instrument in enabling hospitals to learn from each other, to initiate performance improvements and catalyze knowledge-sharing.

DESCRIPTIVE STUDIES

The findings presented in Chapter 2 are important for hospital management and surgical teams, since they clearly suggest that improving the utilization of OR time should be focused on reducing the amount of underused (empty) OR time at the end of the day. This performance indicator has the strongest influence on raw utilization (%), followed by late start and turnover time. The relationships between the three ‘nonoperative’ time indicators were negligible. The impact of the partial indirect effects of ‘nonoperative’ time on utilization were statistically significant, but relatively small.

Based on this study, late start, turnover time and underused time were ‘stand-alone’ aspects with an important direct influence on raw utilization and only a minor influence on each other. We were unable to verify the earlier reported ‘trickle down’ effect7, caused by late start and resulting in an increased delay as the day progresses. Potential solutions and interventions to address the issue of underused OR time are: improving the prediction of the total procedure time of surgical cases; altering the sequencing of scheduled operations and altering patient cancellation policies.

230

12

Chapter 3 identified that in a university hospital environment a quarter of Total Procedure Time (TPT) is Anesthesia-Controlled Time (ACT). As a result, it was concluded that grossing up the predicted Surgeon-Controlled Time (SCT) by 33% to account for the expected ACT, can improve the prediction of TPT if this methodology is adopted. This confirms that employing a fixed time period for ACT (e.g. 20 minutes), is unsuitable because like SCT, ACT is subject to variability. The results affirm that ACT is a considerable part of TPT, which should be scheduled just as realistically as SCT. Robust OR schedules need to anticipate SCT as well as ACT. ACT should be predicted apart from SCT, as a separate time period instead of one combined predicted time period for TPT. More accurate prediction rules may lead to less under- and over-utilized OR time and a reduction of case cancellations8,9.

Thirty-three percent is a higher proportion than reported in earlier research10. This recommendation will improve OR scheduling, which might result in the reduction of under- and over-utilized OR time as well as a reduced amount of case cancellations, and therefore in more efficient use of limited OR resources8,9,11. Recently one Dutch UMC, the Academic Medical Center (AMC) in Amsterdam, adopted a system of scheduling ACT based on predetermined time frames per anesthetic technique. These time frames were differentiated according to the quantity of anesthesia monitoring needed and the complexity of the patient. The implementation of this new scheduling method started at the end of 2012 and we have currently conducted further research to assess the value and effects of this methodology in practice (Chapter 6).

Chapter 4 assessed the effect of surgeons and anesthesiologists on the prediction of OR time. Previous work showed the importance of the surgeons and the surgical team for prediction of OR time11-13. However, this study is the first to show the actual effect sizes of surgeons and anesthesiologists on OR time in a multivariate model corrected for various known predictors. The actual effect of individual surgeons and anesthesiologists is rather small. The overall effect of the first surgeon could explain only 2.7% of the total variation in OR time. Nonetheless, including the individual members of the surgical team in the prediction model improved its accuracy and reduced the over- and underestimation of OR time.

INTERVENTIONAL STUDIES

To identify whether the Dutch OR benchmarking collaborative has led to sustainable improvements in OR management in time, several quasi-experimental studies were performed on specific relevant subjects. To begin, a ‘late start’ or first-case tardiness (Chapter 5) is still a common source of frustration in the OR department. On an overall level of eight UMCs in the Netherlands, 43% of all first operations start at least 5 minutes later than scheduled and 425,612 minutes are lost due to this annually, which has a respectable economic impact.

231

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

On the other hand, the results show that on an overall level of all UMCs first-case tardiness has decreased since 2005 and four centers implemented successful interventions to reduce tardiness. These UMCs showed a stepwise reduction in variation of first-case tardiness, in other words a decrease in IQR during the years, which indicates an organizational learning effect14. ANOVA with contrast analysis shows that a marked change occurred at the time of the intervention, which indicates the success of their interventions. First-case tardiness occurs on a daily basis in Dutch UMCs and this has a sizeable impact on OR efficiency. Yet, this study shows that benchmarking can help to overcome this by exchanging best practices and prevent ‘reinventing the wheel’ through organized learning and networking. In accordance with De Korne et al. 1 this research further supports the idea that benchmarking is highly dependent on social processes and a learning environment parallel to a structured and rational process of sharing performance data. Transfer of knowledge is one of the major targets of the Dutch OR Benchmarking collaborative. During the two-monthly organized multidisciplinary focus-group study meetings and the yearly national invitational conference, targets and goal setting are a matter of discussion and presentation. The overall data presentation is complemented by best practices from different hospitals. Thus, knowledge transfer is performed according to two routes: data analysis and best practice sharing. Overall, this study shows that benchmarking can be applied to identify and measure the effectiveness of interventions to reduce first-case tardiness in a university hospital OR environment.

In 2012, in AMC Amsterdam the OR management decided to implement a new strategy regarding realistic scheduling (Chapter 6). This new method comprised of developing predetermined time frames per anesthetic technique based on historical data of the actual time needed for anesthesia induction and emergence. In total seven so-called ‘anesthesia scheduling packages’ (0 – 6) were established. Several options based on the quantity of anesthesia monitoring (e.g. intubation, arterial line, central line) and the complexity of the patient were differentiated in time within each package: e.g. general anesthesia with tube 30 minutes or awake fiber-optic intubation with epidural 80 minutes.

The most prominent results to emerge from this study are the reductions in prediction errors as well as in the number of case cancellations since the implementation of this new scheduling method specifically for anesthesia-controlled time. Simultaneously, the number of cases performed, increased along with an increase of mean total procedure time.

These findings provide important implications with respect to OR scheduling in a university hospital setting, since they affirm that anesthesia time is a considerable component of total procedure time and should be scheduled just as realistically as and separate from surgeon-controlled time. Scheduling the two major components of a procedure (ACT as well as SCT) more accurately, results in less case cancellations and lower prediction errors. This may lead to more patient satisfaction and a more efficient use of limited and expensive OR resources.

In recent years, there has been an interesting development in one of the Dutch UMCs,

232

12

i.e. Radboud UMC in Nijmegen, regarding operating room scheduling, which received a lot of attention in the OR benchmarking network. This intervention comprised of the implementation of ‘cross-functional OR scheduling teams’ (CFTs) for every surgical department. Every CFT is headed by an anesthesiologist and also includes a surgeon, a scheduler, an OR nurse, an anesthesia nurse, a recovery room nurse, and a nurse from the ward. One a week there is a team meeting to discuss the OR schedule of the next week and to evaluate the OR performance of the previous week, in terms of utilization, case cancellations and other factors interfering with ‘smooth’ planning. The CFT examines the complete OR program, day by day and members inform their colleagues regarding all relevant issues needed for optimal planning and safety. The CFT was given full ‘mandate’ (or ‘authorization’) by the Head of the Department of Operating Rooms and by the Head of the Department of Anesthesiology, to make operational decisions regarding the OR schedule and to make alterations to the submitted OR schedule (e.g. change the order of cases or to not approve of a submitted schedule when the scheduled time exceeds the 8h OR block time allocated to a specific surgical department).

Three studies were conducted on the subject of CFTs: a single-center qualitative case study (Chapter 7), a single-center study with a longitudinal quantitative research design (Chapter 8), and a multi-center study with a quasi-experimental time-series design (Chapter 9). The findings of these three studies highlight the importance of team-based approaches and the need to improve multidisciplinary collaboration between healthcare professionals. The best-performing teams could identify bottlenecks at an earlier stage and were able to solve these bottlenecks. Reduction of uncertainties – by means of optimizing multidisciplinary collaboration – will improve OR scheduling (Harders et al., 2006). In other words, CFTs are assumed to have a self-regulating capacity to identify bottlenecks and to improve continuity. The teams gained insight into their performance using several performance indicators. Consequently, through collaboration, these teams could both control and learn (Chapter 7).

Moreover, our research identified a gradual improvement in OR utilization (2005 – 2011) for two specific surgical departments in Radboud UMC. Results showed a significant reduction in variation – a decrease of interquartile range during the years – of utilization since the implementation of CFTs and a significant increase in mean raw utilization every year. The stepwise reduction of variation indicates an organizational learning effect and more consistency in OR scheduling (‘in control’). The increase of mean raw utilization during the years and the reduction of uncertainty are indicators of more efficient utilization of scarce OR time (Chapter 8). Furthermore, the multicenter study strengthened the idea that multidisciplinary collaboration in CFTs during the perioperative phase has a positive influence on OR scheduling and the use of OR time. Radboud UMC had the highest median raw utilization – 94% versus 85% group median of six other UMCs – during the years 2005 up to and including 2013. An interesting additional finding is that other national databases, concerning mortality rates, support the

233

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

idea that introducing CFTs can be important for improving the quality and safety of care, since Radboud UMC showed the lowest mortality and lowest complication rates (Chapter 9).

The last interventional study (Chapter 10) in this thesis actually covers two important topics: OR capacity for emergency surgery and the value of computer simulation modeling in the complex environment of the OR department. The study showed that in daily practice a dedicated OR for emergency cases is preferred over the approach of evenly reserving capacity for emergency surgery in all elective ORs, in performance terms of raw utilization (%), ‘overtime’ and the number of ORs running late. Moreover, the additional data concerning case cancellations indicates that a dedicated emergency OR has the benefit of less case cancellations. The study assessed the change in efficiency parameters of the problem that was simulated by Wullink et al.15. The results of their specific simulation study led to closing of the emergency OR in the Erasmus MC and evenly allocating emergency capacity in all elective ORs. However, the results of this recent study are partially in contrast with the simulated results: in retrospect, overtime significantly increased after closing the emergency OR. These recent results were based on the empirical data. OR utilization did increase, nevertheless, this increase was lower than the increase in utilization found in both control UMCs without the specific intervention of closing the emergency ORs. Therefore, these recent results do not support the earlier published conclusions5, 6, 11, 12 that distributing free OR capacity for emergency surgery evenly over all elective ORs performs better than dedicated emergency ORs on measures regarding the efficient use of scarce OR time.

Furthermore, the results corroborate the idea of Fone et al.16 that computational modeling experiments are important to support evidence-based policy making in hospital care but they are not able to reflect all complexities of organizations, such as OR departments, and the people working there. In other words, the theory is right but practice is different. A reasonable approach for future interventions in the multifaceted OR environment could be to follow-up policy decisions that are based on simulation modeling results with a robust controlled time-series research design using empirical data, in which the change in parameters is carefully assessed. To further support ‘Evidence-Based OR Management’, every intervention should be followed-up by an evaluation (‘check’) and, if needed, followed by adjustments and new actions (‘act’), in line with the well-known Deming’s cycle.

234

12

REFERENCES

1. De Korne DF, Sol KJ, van Wijngaarden JD, et al. Evaluation of an international benchmarking initiative in nine eye hospitals. Health Care Manage Rev 2010;35:23-35.

2. De Korne DF, van Wijngaarden JD, Sol KJ, et al. Hospital benchmarking: are U.S. eye hospitals ready? Health Care Manage Rev 2012;37:187-98.

3. Fixler T, Wright JG. Identification and use of operating room efficiency indicators: the problem of definition. Can J Surg 2013;56:224-6.

4. Kazemier G, Van Veen-Berkx E. Comment on “Identification and use of operating room efficiency indicators: the problem of definition”. Can J Surg 2013;56:E103-4.

5. APQC. Benchmarking Code of Conduct. Guidelines and Ethics for Benchmarkers. 2008.

6. Askim J, Johnsen A, Christophersen KA. Factors behind organizational learning from benchmarking: Experiences from norwegian municipal benchmarking networks. J Publ Adm Res Theor 2008;18:297-320.





11. Dexter F, Dexter EU, Masursky D, Nussmeier NA. Systematic review of general thoracic surgery articles to identify predictors of operating room case durations. Anesthesia and analgesia 2008;106:1232-41, table of contents.

12. Strum DP, Sampson AR, May JH, Vargas LG. Surgeon and type of anesthesia predict variability in surgical procedure times. In: Anesthesiology; 2000:1454-66.

13. Wright IH, Kooperberg C, Bonar BA, Bashein G. Statistical modeling to predict elective surgery time. Comparison with a computer scheduling system and surgeon-provided estimates. In: Anesthesiology; 1996:1235-45.




235

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

General Limitations

This thesis aims to answer the question whether a nationwide long-term benchmarking collaborative of the operating room departments of all eight university medical centers in the Netherlands has led to improvements in overall OR management. Using exploratory, descriptive and experimental (or interventional) research, this main question can be answered with “yes”. Nevertheless, a few general limitations should be addressed.

The question remains on the topic of the generalizability of study results. The OR benchmark data were gathered in tertiary referral centers only, and therefore generalization of the findings to general hospitals may be limited. The complexity of surgical cases as well as their duration is usually greater than in general hospitals. This was also validated in Chapter 3: the mean (SD) Total Procedure Time of 158 (119) minutes and the median of 124 minutes reflect that the complexity of procedures is potentially greater than in other facilities. This level of complexity of the patient case mix in UMCs can make it more difficult to accurately predict their duration and hamper efficient OR scheduling. Uncertainty, variability and length in the duration of surgery contribute to the difficulty of scheduling1, 2, which may lead to either much underutilized time or unwanted overtime at the end of the day.

One can imagine that in general hospitals with less complex patients, shorter case durations and the attendant reduced variability, case durations can be more accurately predicted. This, in turn will result in more effective scheduling with efficient use of OR resources3 (with less underutilized time and less overtime). One can also imagine that in general hospitals with smaller OR facilities (e.g. with a total of up to 10 ORs) turnover times can be shorter than in UMCs with large OR facilities (20 or more ORs). Smaller facilities deal with a shorter patient transport time from ward to the holding area and from the holding area to the operating room.

In addition, the complexity of tertiary surgical cases and their relatively long total procedure time was also validated in Chapter 6: the mean (SD) total procedure time of 186 minutes (SD 127) reflects that the complexity of procedures is potentially greater than in other hospitals. In a university hospital setting a minimum of 25% up to 30% of total procedure time is engaged by anesthesiologists4. Undoubtedly, this proportion will be smaller in general hospitals. The anesthesia scheduling packages as developed in AMC Amsterdam might be a far too extensive method for other, smaller hospitals organized according to a ‘focused factory’ model, which is characterized by a uniform approach for each patient population segment.

Another issue with the research discussed in this thesis and other past studies related to OR scheduling and OR efficiency lies in the way data are collected5. In the Netherlands, the OR departments of all eight UMCs established a benchmarking collaborative in 2004, continuing to this day. Each UMC submits the data records of all surgical cases performed to a central OR Benchmark database. All data are prospectively, electronically entered in real-time by the OR nursing staff into a Hospital Information System per UMC and subsequently

236

12

confirmed by the surgeon and anesthesiologist in charge. The individual databases of each of the eight UMCs are originally intended for administrative and managerial purposes. We acknowledge the potential, virtually unavoidable biases stemming from this data collection source (administrative/nursing database) and agree with Overdyk’s5 remark that it might even be impossible to exclude bias when data collection depends on human individuals instead of automatic electronic time recording systems.

The estimated magnitude of this ‘human bias’ in this longitudinal OR benchmark study is considered to have a small impact because of the long-term stable nature of data capture. It involves repeated and continuous measurement of the same parameters over a long period of time. In this respect, we have assessed the OR benchmark data and found that parameters over all these years (2005-2015 and still enduring) either show a consistent picture over the years, a gradual increase or a gradual decrease. Furthermore, the differences between the UMCs also show a consistent picture, which does not indicate that human bias is of imperative size.

With reference to the central OR Benchmark database and the variables measured in our studies, a major concern of readers could be the distribution of the data and the manner of statistical testing. Data showed a positively-skewed lognormal distribution, thus, the assumption of normality was dishonored. However, ANOVA is considered a robust test against the normality assumption, particularly with large sample sizes (N ≥ 1,000), which was the case in this study. This is particularly true for larger sample sizes, since the sampling distributions then have weaker dependence on the shape of the population distribution6, 7,

8, 9. In addition, Kruskal-Wallis one-way analysis of variance showed the same results and therefore one-way ANOVA with contrasts was further applied in this study to compare more than two groups. Concerning linear regression analysis, normality of data is not a principal assumption. Normality of the error distribution is a principal assumption, which justifies the use of linear regression, yet again, it is not imperative for large sample sizes (N ≥ 1,000), which was the case in our studies6, 7, 8, 9.

A final limitation regarding the interventional studies conducted, is that within a health care system, particularly in a complex and dynamic environment such as the OR, multiple changes occur during any given period. The evaluation of quality improvements (interventions), frequently rely on weak “before-after” designs10. These “other” changes might have produced the preferred improvements, instead of the specific intervention. One way to minimize this possibility, is to consider multiple time periods in a time-series design as used in this research. The ANOVA with three contrasts analysis, applied in several studies described in this thesis, conveys the extent of background variation and also indicates the extent to which any trend toward improvement may have been present prior to the intervention10.

237

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

REFERENCES






6. Larson MG. Analysis of variance. Circulation 2008;117:115-21.

7. Neill J. Writing Up An ANOVA Analysis. Center for Applied Psychology, University of Canberra 2007.




238

12

General Discussion and Future Perspectives for Research

Benchmarking OR Departments in the NetherlandsBenchmarking has often been approached as a competitive activity resulting in rankings and with a focus on creating competition between participants as driver for improvement. However, this study (Chapter 1) clearly shows the advantages of a more collaborative approach. An important difference between public reporting and reporting arranged in this Dutch benchmarking collaborative is the fact that the performance as well as rankings are not publicly available elsewhere than to the eight participating UMCs. When information is publicly and freely available, it will be more difficult to build a relation of trust.

From the very first start, the initiators of the Dutch OR benchmarking collaborative as described in this study consistently and literally have avoided ‘naming and shaming’ through publishing and vertical ranking of the eight UMCs, regarding the performance indicators measured. Lots of attention has been given to honest assessment and avoiding to compare apples and oranges. The physical, organizational characteristics and structure of all participating OR departments can be very different from one another. Contingency theory claims there is not “one best way for organizing” because this is subject to the internal and external conditions of every organization1-3. Differences in organizational characteristics derive from differences in organizational conditions. Therefore, performance indicators used for benchmarking should take into account these differences, to avoid inaccurate interpretation of observed differences between organizations and to accomplish an honest comparison.

The character of benchmarks using DEAs is essentially different from the character of the Dutch OR Benchmarking Collaborative since it was initiated by the eight university hospitals themselves and not by a third external party. Moreover, data is derived from the local Hospital Information Systems, which are used for daily registration practices. The Dutch OR Benchmarking Collaborative is a ‘self-led’ and voluntary collaboration with its own budget (paid for by the eight hospitals themselves). OR benchmark data is merely used by the participants and not by policy makers, the government or regulatory offices.

Another foundation of the collaborative benchmark described in this study, is the pursuit to learn from the organizational differences in structure, process designs, methods and performance. These differences can be a source of learning as they allow practitioners to compare relations between organizational characteristics and performance, especially in informal settings and networking. These differences also offer every participating OR department the opportunity to engage their own quality improvement pathway. Improvement starts with quantitative analyses and therefore performance indicators should be SMART. In this collaborative the interorganizational or ‘joint learning process’ is more important than ranking participants or to identify ‘the best practice’. The OR departments of the eight UMCs

239

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

are all providing the same healthcare product: perioperative care in a university hospital setting. It is important to gain insight into managing and controlling this process as well as insight into performance differences, to realize the ‘best fit’ for each OR department.

Nevertheless, the OR benchmark collaborative could learn from e.g. the IHI breakthrough series approach to develop a more structured PDSA-approach. Specifically with regard to the commitment of the participants during study meetings and the (learning) activities in between physical meetings. When a healthcare professional decides to participate in a breakthrough series, he commits to participate actively throughout the limited collaborative period. In general, this period is limited to 6 - 18 months, which is supposed to drive change. In between physical meetings, teams are expected to implement changes in their own organization and it is mandatory to share implementation experiences with each other for collective learning through conference calls or digital (internet) platforms4-6. Clearly, this kind of ‘stable commitment’ through continuous participation was not established in the OR benchmark collaborative in this recent study. Healthcare professionals that visit a focus group study meeting are not permanent delegates since they are not obliged to visit the following meetings. The responsibility for improvement was kept an individual responsibility of each single UMC and not a collaborative responsibility. Future research should therefore concentrate on the investigation of the relation between benchmarking as a managerial instrument and the actual performance improvements realized through benchmarking in the local UMC’s.

DESCRIPTIVE STUDIES

The study findings in Chapter 2 suggest that OR utilization can be improved by focusing on the reduction of underused OR time at the end of the day. This might be possible by altering patient cancellation policies7, 8, 9, 10. A practice applied in many Dutch (university) hospitals is a ‘zero tolerance for overtime’ policy, because OR management presumes it is more economically profitable to finish the daily OR caseloads during ‘regular’ hours than to create overtime9, 10. A consequence of this policy may be that a patient scheduled in the final allocated hours (or late afternoon) will be cancelled last-minute to avoid overtime. This leads to immaterial damage concerning postponed or cancelled patients and to financial losses for the hospital concerning under-utilization of scarce OR capacity. Because all OR personnel in Dutch UMCs are contracted and paid for at least 8 hours on each day worked, underutilized time leads to economic losses for the hospital due to these fixed labor costs. Tessler10 and Stepaniak9, however, showed in their previous work that it is more cost-effective to proceed with an operation after regular hours than to cancel this operation. Overtime does have a financial effect owing to the payment of overtime wages beyond the regular rate for 36 hours a week (in Dutch UMCs). Working overtime can also have a negative influence on job

240

12

satisfaction of registered nurses and is considered a reason to change their employment status11, 12. To better absorb the consequences of underutilized time as well as overtime, one option could be to employ OR personnel on a flexible basis adjusted to patient needs, as suggested in previous research13-16.

The direct and indirect effects of ‘non-operative’ time on raw utilization are worthwhile studying because former research concluded that late start can cause a ‘trickle down’ effect resulting in an increased delay (of e.g. turnover time) as the day progresses, potentially affecting the rest of the scheduled patients17. This research, however, implicates that the indirect effects of late start through turnover time and underutilized time do not have a major impact on raw utilization. Therefore, these recent results reconfirm several earlier studies that resources spent solely on trying to achieve on time starts of scheduled first cases will not considerably improve OR utilization or productivity18-22, and the ‘trickle down’ effect has not yet been verified13, 21-23. Our study reveals that a late start can be caught up throughout the rest of the OR-day, either during operative time or due to a quicker turnover. Future research should investigate this specific subject to reveal its principles.

The results in Chapter 3 further support the idea that scheduling surgical procedures is a complex process because a procedure entails several elements subject to variability, such as room setup and takedown, patient positioning, prepping and draping, as well as the two principal components surgeon-controlled time and anesthesia-controlled time. Nevertheless, there are even more variability factors that can be of influence, such as patient characteristics (e.g. age, BMI, comorbidities, number of previous operations) as well as OR team member characteristics and their experience (attending, fellow, resident, trainee and experience in years). Even for experienced anesthesiologists, it is often difficult to predict how long the anesthetic induction for a specific patient will take24. Factors such as ASA physical status, anesthetic technique (e.g. monitors, lines, pain management procedures), working with trainees and residents in a teaching setting and the surgical procedure, have shown to affect ACT and SCT significantly25,24,26,27. Because the central OR Benchmark database was not designed to register all of these variability factors, this study could not investigate their specific impact.

Future studies, which take more variability factors into account, will need to be undertaken. On individual hospital level these factors are partially available. It would also be interesting to compare SCT among surgeons regarding the same procedure, as well as ACT among anesthesiologists regarding the same anesthetic technique. Using historical times per surgeon and per procedure to schedule SCT is not new28-33, using historical times per anesthesiologist, however, is not common in the Netherlands. Recently one Dutch UMC, the Academic Medical Center (AMC) in Amsterdam, adopted a system of scheduling ACT based on historical times per anesthesiologist and per anesthetic technique. The implementation of this new scheduling method started at the end of 2012 and we currently conducted further research to assess the

241

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

value and effects of this methodology in practice (Chapter 6).Chapter 4 assessed the effect of surgeons and anesthesiologists on the prediction of

OR time. Previous work showed the importance of the surgeons and the surgical team for prediction of OR time34-36. However, this study is the first to show the actual effect sizes of surgeons and anesthesiologists on OR time in a multivariate model corrected for various known predictors. The actual effect of individual surgeons and anesthesiologists is rather small. The overall effect of the first surgeon could explain only 2.7% of the total variation in OR time. Nonetheless, including the individual members of the surgical team in the prediction model improved its accuracy and reduced the over- and underestimation of OR time.

The final model contained significant interaction terms between the surgeon and the type of procedure. This indicates that the effect of the surgeon on OR time is lower or higher for different types of procedures. Several explanations can be given for this significant interaction. First, Strum et al. mentioned that surgeons consistently work in a different pace (tempo): the work rate effect35. Therefore, differences between surgeons increase proportionately with longer procedures and thus increase the variability in OR time. Second, several studies have shown that the experience of the surgeon or the surgical team influences the duration of the OR time37, 38. These studies illustrate that increased experience (expressed as times performed the procedure) lowers the duration of surgery. Therefore, surgeons with less experience with a certain type of procedure are likely to show more variability in their OR time. At last, these data originate from a university medical center where oncologic procedures were performed by specific surgeons. For oncologic procedures, the discrepancy between procedure times can be high due to incorrect pre-operative tumor staging or conversion of the planned procedures. For example, during laparoscopic tumor resections, conversion rates to open procedures can be as high as 20%39. Incorrect preoperative staging may reveal inoperable oncology during surgery. In that case, the predicted duration will be much longer than the actual time and leads to an increased variability in OR time.

Future research on OR scheduling should involve the following topics. First, investigate the effect of surgeons and anesthesiologists by a separate analysis of anesthesia-controlled time and surgeon-controlled time. Second, separate variance components have to be determined for each surgeon individually, independent if he or she is the first or second surgeon. At last, the predictive model should incorporate random slopes for known predictive factors (i.e. estimated time, patient and procedure characteristics).

242

12

INTERVENTIONAL STUDIES

Four UMCs implemented interventions to reduce first-case tardiness (Chapter 5) and effectuated significant results. However, the ANOVA with contrast results of specifically one UMC demonstrated that the trend toward improvement might have been present prior to the intervention. This suggests that the original high value of tardiness of this specific UMC at the starting point might be an important determinant for improvement. A high sense of urgency is a critical success factor for a change process to succeed40. Purely based on the original values of tardiness from the other three UMCs, these centers had less sense of urgency and fewer room to improve first-case tardiness; nevertheless, they did and also succeeded significantly. The specific UMC showed the highest relative improvement because of their lower original value and thus having more room to improve than the other centers.

Although recent studies indicated that first-case tardiness does not affect OR efficiency22,

41, 42 and the ‘trickle down’ effect has been argued against22, 21, 18, 23 (see also Chapter 2), first-case tardiness remains of interest because it continues to be perceived as a key performance indicator of inefficiency in the OR43. Moreover, this can be confirmed in the OR practice of all eight UMCs in the Netherlands, as 28% of the variation in raw utilization was explained by the variation in first-case tardiness in the current study. Also other fundamental elements might be influenced by it in a negative manner. Patient satisfaction may be reduced if operations are delayed beyond their scheduled start times, particularly if patients who had to fast are kept waiting for several hours44, 43, 45, 17. Furthermore, delays are a source of frustration for health care professionals and, although, time saved by reducing first-case tardiness cannot be accommodated with extra operations, the time saved is still time that can be used for other purposes43, 46. The multidisciplinary focus group in this study corroborated that starting on time means less rush at the beginning and potentially throughout the day; and rushing has been identified as one of the factors that lead to an unsafe working environment17, 47. In this context, the outcomes of this study may contribute to the improvement of overall operating room practice.

Reducing first-case tardiness and increasing the proportion of on-time starts is merely one aspect of efficient use of OR capacity. In ORs, inefficiencies can occur before, during, between and after cases48. Further research is required considering the additional performance indicators in this nationwide multicenter OR Benchmark database such as turnover time, under-utilized OR time, over-utilized OR time, case cancellations and prediction errors (see all other Chapters).

The research described in Chapter 6 shows the possibility to improve OR scheduling with a new method entitled ‘anesthesia scheduling packages’ as developed in AMC Amsterdam. The number of patient case cancellations as well as prediction errors decreased. This method differentiates in time by taking into account the quantity of anesthesia monitoring as well as

243

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

the complexity of the patient, in contrast to the common method of assigning a fixed number of minutes for anesthesia. The package is assigned already during the pre-anesthesia check-up where medical complexities of the patient related to anesthesia and surgery are evaluated.

Additionally, unanticipated results derived from our study, demonstrating the clinical relevance of this research. These are beneficial side effects due to information that came available earlier in the patient process: i.e. the required anesthesia package that is assigned during the pre-anesthesia check-up. Now, anesthesia nurses know exactly which medical equipment and devices have to be assembled and tested at forehand. Correspondingly, anesthesia residents know in advance in which operating room a complex anesthetic technique, like an awake fiberoptic intubation, is scheduled, so they can watch and learn. Moreover, in light of technical skills training, the scheduling and registration of anesthesia packages supplements the clinical training portfolio of residents as well as anesthesiologists.

A final, unanticipated result was the observed improvement in communication between surgeons and anesthesiologists. Because it is now transparent how long anesthesia time will take before the start of every operating room session, surgeons started to take this into account with regard to the complete OR schedule: e.g. nowadays, a patient with a scheduled peripheral nerve block is placed second instead of first on the schedule because anesthesia time will take approximately 60 minutes. Additionally, surgeons suggested to start with arterial line placement of the second patient while the first patient is still in the OR, which is called ‘parallel processing’. This further supports the idea of Friedman et al.49 that starting the anesthesia process in the preoperative holding area, including intravenous sedation and local anesthesia administration, can realize a significant decrease in the amount of in-room time spent.

Nevertheless, dividing OR scheduling into the two main components SCT and ACT, pays no specific attention to the time interval used for positioning, prepping and draping prior to incision. This preparation time is now incorporated in SCT. Especially in a university hospital environment with complex surgical procedures this can take a considerable amount of time and can, therefore, be of influence on the prediction of total procedure time. It would be interesting for future research to focus on differentiating this preparation time per surgical procedures in order to establish whether this would create more accurate predictions.

The implementation of ‘cross-functional OR scheduling teams’ (CFTs) in Radboud UMC was evaluated with three separate research projects (Chapters 7, 8 and 9). This research revealed that the key differences between high-performing and low-performing teams are common goal setting, single-loop and double-loop learning, which are essential for continuous improvement. In particular double-loop learning and control mandates were important in the high-performing teams, that were able to accommodate multidisciplinary professionals and therefore improved continuously during the study period. Cohesion, openness and feedback are indirectly essential to improving performance. The low-performing teams did not hold their

244

12

members accountable for their actions, and therefore, the learning circle was not completed. Also, the team members’ self-interest regularly took priority over the public’s interest.

The high-performing CFTs have learned how to deal with interferences, and how to improve their services continuously through better collaboration and better use of control mechanisms (i.e. feedback loops, policy adjustment). An organization that is able to achieve a range of objectives, despite variability and interferences, is said to be ‘in control’50. Abandoning the so-called ‘functional silos’ and improving multidisciplinary collaboration in the perioperative phase contributes to an efficient OR schedule. An OR schedule, well prepared by a CFT, will reduce the number of cancellations and improve the prediction process for the next schedule. This is expected to keep waiting lists for ORs as short as possible51.

To fully understand the influence of CFTs on OR performance, future research is required at all surgical departments to discover if the results are as promising as those presented in our three recent studies. Also, the analysis of several additional separate performance indicators, e.g. first-case tardiness, turnover time between cases, under- and over-utilized time as well as case cancellations, can identify areas of further improvement in relation to CFTs.

Finally, it is not yet clear whether multidisciplinary collaboration in CFTs also leads to better quality of care51-53 since we have not yet explicitly investigated this. Despite the encouraging numbers in the Adult Cardiac Surgery Database54, which show that Radboud UMC has the lowest mortality, and lowest complication rates in comparison with the control group55, this question remains unanswered and should be the topic of future research.

In the last interventional study (Chapter 10) of this thesis the approaches for reserving OR capacity for emergency surgery were discussed, as well as the value of computer simulation modelling in complex hospital environments, such as OR departments. The earlier published conclusions5, 6, 11, 12 which led to closing the emergency OR in the Erasmus MC, were correctly based on simulated results. One major drawback of this approach is that simulation results do not always reflect the discouraging reality, which was the case in this recent study. Earlier findings in favor of a dedicated emergency OR were based on either simulation modeling56,

57, or empirical data analyzed with a straightforward “before-after” approach58-61. “Before-after” designs suffer two disadvantages: first, background factors can produce large fluctuations in processes or outcomes of interest unrelatedly to the specific intervention that is studied; second, in time, multiple changes occur within a health care system or its socioeconomic environment and one or more of these other changes might have produced the preferred improvements62.

If, in this recent study, merely a “before-after” design was applied, two out of the three performance variables were confirmed, corresponding to the simulated results. Nevertheless, by strengthening the research design with two control UMCs, these effects turned out to be debatable. There are two ways to minimize the drawbacks of simple “before-after” designs,

245

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

i.e. a time-series design with multiple time periods and a controlled before-after design, in which the same measurements occur in one or more hospitals that did not implement the intervention of interest. These two designs convey the extent of background variation. In the current study, both designs were combined, which resulted in applying a quasi-experimental controlled time-series design (ANOVA with contrasts, as well as two control UMCs).

Simulation models as referred to earlier63, 64 described that the model was an ‘abbreviated’ version of reality. Wullink etal.64 specifically mention that “implementation of the policy by which emergency capacity is reserved in all elective ORs, requires all stakeholders on the OR to strictly adhere to the policy”. In practice, not all surgical departments reserved free OR capacity for emergency surgery in their elective program as agreed upon. This hampered the performance of the approach that evenly allocates capacity for emergency surgery to all elective ORs considerably, as this model is most dependent on the collaboration and commitment of all surgical departments. Future research on the outcomes of model implementation in health care settings such as the multidimensional OR environment, is required to contribute to closing the gap between operations research and organizational practices65, 66. The findings of this study do not support strong recommendations to close dedicated emergency ORs, as was recommended earlier based on simulated results, since information on the outcomes of model implementation is missing.

246

12

REFERENCES

1. Burns T, Stalker GM. The management of innovation. London: Tavistock; 1961.

2. Lawrence PR, Lorsch JW. Organization and environment: managing differentation and integration. Homewood (Ill.): Irwin; 1969.

3. Mintzberg H. The structuring of organisations. New York: Prentice-Hall; 1979.

4. Kilo CM. A framework for collaborative improvement: lessons from the Institute for Healthcare Improvement’s Breakthrough Series. Qual Manag Health Care 1998;6:1-13.

5. Kilo CM. Improving care through collaboration. Pediatrics 1999;103:384-93.

6. Nembhard IM. All teach, all learn, all improve?: the role of interorganizational learning in quality improvement collaboratives. Health Care Manage Rev 2012;37:154-64.




10. Tessler MJ, Kleiman SJ, Huberman MM. A “zero tolerance for overtime” increases surgical per case costs. Can J Anaesth 1997;44:1036-41.

11. Shader K, Broome ME, Broome CD, West ME, Nash M. Factors influencing satisfaction and anticipated turnover for nurses in an academic medical center. J Nurs Adm 2001;31:210-6.

12. Strachota E, Normandin P, O’Brien N, Clary M, Krukow B. Reasons registered nurses leave or change employment status. J Nurs Adm 2003;33:111-7.


14. Dexter F, Macario A, Manberg PJ, Lubarsky DA. Computer simulation to determine how rapid anesthetic recovery protocols to decrease the time for emergence or increase the phase I postanesthesia care unit bypass rate affect staffing of an ambulatory surgery center. Anesth Analg 1999;88:1053-63.

15. Macario A, Dexter F. Effect of compensation and patient scheduling on OR labor costs. AORN J 2000;71:860, 3-9.



18. Dexter EU, Dexter F, Masursky D, Garver MP, Nussmeier NA. Both bias and lack of knowledge influence organizational focus on first case of the day starts. Anesth Analg 2009;108:1257-61.

19. Dexter F. Is time on first-case starts well spent? OR Manager 2010;26:22-4.

20. Mazzei WJ. Operating room start times and turnover times in a university hospital. J Clin Anesth 1994;6:405-8.

21. McIntosh C, Dexter F, Epstein RH. The impact of service-specific staffing, case scheduling, turnovers, and first-case starts on anesthesia group and operating room productivity: a tutorial using data from an Australian hospital. Anesth Analg 2006;103:1499-516.

247

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed






27. Urman RD, Sarin P, Mitani A, Philip B, Eappen S. Presence of anesthesia resident trainees in day surgery unit has mixed effects on operating room efficiency measures. Ochsner J;12:25-9.




31. Dexter F, Ledolter J, Tiwari V, Epstein RH. Value of a scheduled duration quantified in terms of equivalent numbers of historical cases. Anesth Analg 2013;117:205-10.

32. Dexter F, Ledolter J. Bayesian prediction bounds and comparisons of operating room times even for procedures with few or no historic data. Anesthesiology 2005;103:1259-167.

33. Dexter F, Traub RD, Fleisher LA, Rock P. What sample sizes are required for pooling surgical case durations among facilities to decrease the incidence of procedures with little historical data? Anesthesiology 2002;96:1230-6.

34. Dexter F, Dexter EU, Masursky D, Nussmeier NA. Systematic review of general thoracic surgery articles to identify predictors of operating room case durations. Anesthesia and analgesia 2008;106:1232-41, table of contents.

35. Strum DP, Sampson AR, May JH, Vargas LG. Surgeon and type of anesthesia predict variability in surgical procedure times. In: Anesthesiology; 2000:1454-66.

36. Wright IH, Kooperberg C, Bonar BA, Bashein G. Statistical modeling to predict elective surgery time. Comparison with a computer scheduling system and surgeon-provided estimates. In: Anesthesiology; 1996:1235-45.

37. Ballantyne GH, Ewing D, Capella RF, et al. The learning curve measured by operating times for laparoscopic and open gastric bypass: roles of surgeon's experience, institutional experience, body mass index and fellowship training. In: Obes Surg; 2005:172-82.

38. Xu R, Carty MJ, Orgill DP, Lipsitz SR, Duclos A. The teaming curve: a longitudinal study of the influence of surgical team familiarity on operative time. In: Ann Surg; 2013:953-7.

39. Thome MA, Ehrlich D, Koesters R, et al. The point of conversion in laparoscopic colonic surgery affects the oncologic outcome in an experimental rat model. In: Surg Endosc; 2008:1988-94.

40. Kotter JP. Leading change. Boston, Mass.: Harvard Business Review Press; 2012.

41. Macario A. Are your hospital operating rooms “efficient”? A scoring system with eight performance indicators. Anesthesiology 2006;105:237-40.

248

12

42. Macario A. The limitations of using operating room utilisation to allocate surgeons more or less surgical block time in the USA. Anaesthesia 2010;65:548-52.


44. Wachtel RE, Dexter F. Influence of the operating room schedule on tardiness from scheduled start times. Anesth Analg 2009;108:1889-901.


46. Panni MK, Shah SJ, Chavarro C, Rawl M, Wojnarwsky PK, Panni JK. Improving operating room first start efficiency - value of both checklist and a pre-operative facilitator. Acta Anaesthesiol Scand 2013.

47. Greenberg CC, Roth EM, Sheridan TB, et al. Making the operating room of the future safer. Am Surg 2006;72:1102-8; discussion 26-48.


49. Friedman DM, Sokal SM, Chang Y, Berger DL. Increasing operating room efficiency through parallel processing. Ann Surg 2006;243:10-4.

50. Sitter LUd, den Hertog JF, Dankbaar B. From Comlex Organizations with Simple Jobs to Simple Organizations with Complex Jobs. Human Relations 1997;50:497-534.

51. Bitter J, van Veen-Berkx E, Gooszen HG. et al. Multidisciplinary teamwork is an important issue to healthcare professionals. Team Performance Management 2013;19:263-78.

52. Schraagen JM, Schouten T, Smit M, et al. Assessing and improving teamwork in cardiac surgery. Qual Saf Health Care 2010;19:e29.

53 . Pronovost PJ, Freischlag JA. Improving Teamwork to Reduce Mortality. JAMA 2010;15:1721-2.

54 . Siregar S. Safety in cardiac surgery. Enschede: Utrecht University; 2013.

55 . The Netherlands Association for Cardio-Thoracic Surgery. Transparency of outcome indicators for Cardio-Thoracic Surgery. In; 2013:1-6.

56 . Persson MJ, Persson JA. Analysing management policies for operating room planning using simulation. Health Care Manag Sci 2010;13:182-91.

57 . Ferrand Y, Magazine M, Rao U. Comparing two operating-room-allocation policies for elective and emergency surgeries. Proceedings of the 2010 Winter Simulation Conference 2010:2364-74.

58 . Bhattacharyya T, Vrahas MS, Morrison SM, et al. The value of the dedicated orthopaedic trauma operating room. J Trauma 2006;60:1336-40; discussion 40-1.

59 . Heng M, Wright JG. Dedicated operating room for emergency surgery improves access and efficiency. Can J Surg 2013;56:167-74.

60 . Lovett BE, Katchburian MV. Emergency surgery: half a day does make a difference. Ann R Coll Surg Engl 1999;81:62-4.

61 . Lemos D, Nilssen E, Khatiwada B, et al. Dedicated orthopedic trauma theatres: effect on morbidity and mortality in a single trauma centre. Canadian Journal of Surgery 2009;52:87-91.

62 . Shojania KG, Grimshaw JM. Evidence-based quality improvement: the state of the science. Health Aff (Millwood) 2005;24:138-50.

63 . Van Essen JT, Hans EW, Hurink JL, Overberg A. Minimizing the waiting time for emergency surgery. Operations Research for Health Care 2012;1:34-44.

249

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

64 . Wullink G, Van Houdenhoven M, Hans EW, van Oostrum JM, van der Lans M, Kazemier G. Closing emergency operating rooms improves efficiency. J Med Syst 2007;31:543-6.

65 . Fone D, Hollinghurst S, Temple M, et al. Systematic review of the use and value of computer simulation modelling in population health and health care delivery. J Public Health Med 2003;25:325-35.

66 . Rousseau DM. Is There Such a Thing as “Evidence-Based Management”? Academy of Management Review 2006;31:256-69.

250

12

Lessons Learned

This research extends our understanding on the science of OR efficiency. OR efficiency is never a single goal: in this thesis it refers to maximizing the utilization of available OR time, minimizing idle time (or non-productive time) and minimizing costs, reducing patient cancellations, reducing overtime, and last but not least, improving OR scheduling to increase case throughput. To finalize, this research has several practical implications and lessons learned for OR management and surgical teams:

• OR utilization can be improved by focusing on the reduction of empty OR time at the end of day. Potential solutions to address this issue are improving the prediction of total procedure time (patient-in to patient-out of the OR) needed per patient; altering the sequencing of scheduled procedures (schedule short procedures before long procedures to limit variability), and altering patient cancellation policies (abolish the ‘zero tolerance for overtime’ policy)1.

• To capture the consequences of empty OR time or overtime at the end of the day and to optimize the utilization of the available OR capacity, OR personnel should be employed on a flexible basis along the lines of the OR schedule, to better coordinate supply and demand for perioperative care1.

• First-case tardiness or a ‘late start’ is a common source of frustration in the OR. Specific interventions can effectuate a significant reduction in tardiness, such as providing feedback directly when ORs started too late (in person and through internal ‘public’ reporting). The OR nurse or a special transport service has to be responsible for transport of the patient from the holding area to the OR, instead of the anesthesia nurse. Assign a ‘post-call anesthesiologist’ during morning hours to avoid tardiness caused by the fact that one anesthesiologist covers two ORs simultaneously. Improve the utilization of available PACU-beds by allocating the beds in advance to elective patients of the specific surgical specialties that utilize the PACU the most. Form agreements with the ICU department on the release of ICU-beds in the morning and extra temporary beds in case there is no capacity available2.

• Employing a fixed time period for anesthesia time (e.g. 20 minutes), is unsuitable because like surgeon-controlled time, anesthesia-controlled time is subject to variability3. It is difficult for surgeons as well as for anesthesiologists to predict the OR time needed for surgery and anesthesia (induction and emergence) respectively. Grossing up the predicted surgeon-controlled time with 33% to account for the prediction of anesthesia-controlled time, improves the prediction of total procedure time3.

251

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

• Develop predetermined time frames per anesthetic technique based on historical data (median instead of mean due to right skewness) of the actual time needed for anesthesia induction and emergence. Differentiate according to the quantity of anesthesia monitoring needed (e.g. intubation, arterial line, central line) and the complexity of the patient. This will result in less case cancellations and lower prediction errors, which will lead to more patient satisfaction and more efficient use of OR capacity. Moreover, this will improve the OR workflow because anesthesia nurses know exactly which medical equipment and devices they have to assemble and test at forehand. It will also improve communication between surgeons and anesthesiologist with regard to the OR schedule4.

• Estimation accuracy varies comprehensively and scheduling methods relying on surgeon- and procedure-specific historical case time averages are forces to use a very small sample. The typical university hospital OR is characterized by a rich variation in patient case mix, making it challenging to forecast total procedure time. Due to wide dispersion, use the historical median to generate the first estimate, subsequently, the surgeon can adjust that estimate based on patient complexity factors and OR team characteristics (e.g. surgical/anesthesia residents)3.

• Though effect sizes are small, individual surgeons and anesthesiologists influence total procedure time. Taking into account the differences between surgeons and anesthesiologists improves the prediction of total procedure time and may lower the risk of over- and underutilization. It is recommended that scheduling becomes more case-specific and considers the surgeons and anesthesiologists. The more a surgeon performs a certain procedure, the less likely he or she is to deviate from the predicted time (experience is an important source of ‘in-between’ and ‘within-surgeon’ variation). Taking into account the experience of the individual surgeon and anesthesiologist might further improve the accuracy of the estimated total procedure time5.

• Multidisciplinary collaboration in CFTs during the perioperative phase has a positive influence on OR scheduling. It will also reduce variation in the OR process and increase the utilization of OR time. Furthermore, CFTs can be important for improving the quality and safety of care6, 7.

• Computational modeling experiments (‘simulation studies’) are important to support evidence-based policy making in hospital care, but they are not able to reflect all complexities of organizations, such as OR departments, and the people working there. Be aware that the theory may be right but practice is almost always different8.

252

12

• A dedicated OR for emergency cases is preferred over the approach of evenly reserving capacity for emergency surgery in all elective ORs, in performance terms of raw utilization (%), ‘overtime’ and the number of ORs running late. Moreover, a dedicated emergency OR has the benefit of less case cancellations8.

• A collaborative and long-term approach of benchmarking is essential when considering benchmarking activities. It is necessary to build and maintain trust between the participants and for the development of uniform data registration and definitions. This approach also creates the best conditions, like a safe learning and non-judgmental environment, for knowledge-sharing and improvement in practice9.

• Public rankings and ‘ad hoc’ benchmark studies, used for accountability purposes by a ‘third party’, need to be avoided. In analogy to recent discussions on value-based healthcare and outcome measurement, it appears that currently uses outcome indicators are not suitable for ranking due to the considerable influence of random variation and case mix factors on performance differences between hospitals9-12.

• Even though all eight Dutch university medical centers participate in this OR benchmark collaborative, it is important to explore opportunities to learn from other hospitals outside the Dutch university hospital environment, outside the Netherlands or even outside the hospital sector. Otherwise, the participating centers would be ‘navel-gazing’ and they would maintain, despite the fact that eight centers take part, a limited focus. Hence, true innovation may be inhibited9.

253

Sum

mar

y an

d G

ener

al C

oncl

usio

ns, G

ener

al L

imita

tions

, Gen

eral

Dis

cuss

ion

and

Futu

re P

ersp

ectiv

es fo

r Res

earc

h, L

esso

ns L

earn

ed

REFERENCES



3. van Veen-Berkx E, Bitter J, Elkhuizen SG, et al. The influence of anesthesia-controlled time on operating room scheduling in Dutch university medical centres. Can J Anaesth 2014;61:524-32.

4 van Veen-Berkx E, van Dijk MV, Cornelisse DC, Kazemier G, Mokken FC. Scheduling Anesthesia Time Reduces Case Cancellations and Improves Operating Room Workflow in a University Hospital Setting. J Am Coll Surg 2016;223:343-51.”

5 van Eijk RP, van Veen-Berkx E, Kazemier G, Eijkemans MJ. Effect of Individual Surgeons and Anesthesiologists on Operating Room Time. Anesth Analg 2016;123:445-51.

6. Bitter J, van Veen-Berkx E, van Amelsvoort P, Gooszen H. Preoperative cross functional teams improve OR performance. J Health Organ Manag 2015;29:343-52.

7. van Veen-Berkx E, Bitter J, Kazemier G, Scheffer GJ, Gooszen HG. Multidisciplinary teamwork improves use of the operating room: a multicenter study. J Am Coll Surg 2015;220:1070-6.

8. van Veen-Berkx E, Elkhuizen SG, Kuijper B, Kazemier G, Dutch Operating Room Benchmarking C. Dedicated operating room for emergency surgery generates more utilization, less overtime, and less cancellations. Am J Surg 2016;211:122-8.

9. van Veen-Berkx E, de Korne DF, Olivier OS, Bal RA, Kazemier G. Benchmarking Operating Room Departments in the Netherlands: Evaluation of a Benchmarking Collaborative between Eight University Medical Centres. Benchmarking: An International Journal 2016;23(5):1171-92.

10. van Dishoeck AM, Koek MB, Steyerberg EW, van Benthem BH, Vos MC, Lingsma HF. Use of surgical-site infection rates to rank hospital performance across several types of surgery. Br J Surg 2013;100:628-36; discussion 37.

11. Van Dishoeck AM, Lingsma HF, Mackenbach JP, Steyerberg EW. Random variation and rankability of hospitals using outcome indicators. BMJ Qual Saf 2011;20:869-74.

12. van Dishoeck AM, Looman CW, van der Wilden-van Lier EC, Mackenbach JP, Steyerberg EW. Displaying random variation in comparing hospital performance. BMJ Qual Saf 2011;20:651-7.

AppendicesLetters to the Editor / Commentaries

Samenvatting

Acknowledgments

Dankwoord

Publications and Presentations

256256

© 2013 Canadian Medical Association Can J Surg, Vol. 56, No. 5, October 2013 E103

CORRESPONDENCE • CORRESPONDANCE

IDENTIFICATION AND USE OF

OPERATING ROOM EFFICIENCY

INDICATORS: THE PROBLEM OF NOT

PERFORMING THE RIGHT SEARCH

WITHIN PUBMED

Drs. Fixler and Wright1 should becommended for demonstrating thatoperating room (OR) performanceindicator definitions vary in literatureand among children’s hospitals. Unfor-tunately, I do not agree with their con-clusion that the most logical coursewould be for professional associationsto agree upon and develop commonmetrics and definitions. Their conclu-sion is based on a limited review ofpapers that are not always relevant.

First, the Procedural Times Glos-sary has been the leading source forOR definitions since 1997.2 Papersdescribing operational research inORs use this glossary.3 A bibliographyof papers concerning operationalresearch within the OR can be foundonline (http://www .franklindexter .net/bibliography _TOC.htm).

Based on this evidence, I concludethat there are clear definitions formonitoring OR performance indica-tors. An additional conclusion is thathospitals continue to use their owndefinitions. This needs to be solved bysending surgeons, anesthesiologistsand managers of ORs to courseswhere they can learn which indicatorsto use and how to use them.

Fixler and Wright call for us to usethe OR resources in both an efficientand effective way. Here they make amistake. Indeed, monitoring the oper-ational performance of the OR maycontribute to the use of OR resourcesin an efficient way. However, the callto use OR resources in an effective wayis a faulty statement. According to theInstitute of Medicine’s Committee onQuality Health Care in America,effective care “is based on providingservices based on scientific knowledgeto all who could benefit, and refrain-ing from providing services to those

not likely to benefit (avoiding under-use and overuse, respectively).”4 Herethe patient clinical parameters are ofinterest and not, for example, the uti-lization rate of the OR.

In conclusion, performing an accu-rate search in PubMed will show thatthe actual problem of agreed-upondefinitions in literature, as describedby the authors, does not exist.

Pieter Stepaniak, PhDCatharina HospitalOperating Room DepartmentEindhoven, the NetherlandsGdansk Medical UniversityDepartment of General, Endocrine and Transplant Surgery

Gdansk, Poland

Competing interests: None declared.

DOI: 10.1503/cjs.019413

References

1. Fixler T, Wright JG. Identification and useof operating room efficiency indicators: theproblem of definition. Can J Surg 2013;56:224-6.

2. Donham RT, Mazzei WJ, Jones RL. Associ-ation of anesthesia clinical directors' pro ceduraltimes glossary: glossary of times used for sched-uling and monitoring of diagnostic and ther -apeutic procedures. Columbus (OH): Associ-ation of Anesthesia Clinical Directors;1996.

3. Stepaniak PS, Heij C, Buise MP, et al.Bariatric surgery with operating room teamsthat stayed fixed during the day: a multicen-ter study analyzing the effects on patientoutcomes, teamwork and safety climate, andprocedure duration. Anesth Analg 2012;115:1384-92.

4. Institute of Medicine, Committee onQuality Health Care in America. Crossingthe quality chasm: a new health system for the21st century. Washington (DC): NationalAcademic Press; 2001.

COMMENT ON “IDENTIFICATION AND

USE OF OPERATING ROOM

EFFICIENCY INDICATORS: THE

PROBLEM OF DEFINITION”

It was with profound interest that weread the commentary written byTamas Fixler and James G. Wright in

the August 2013 issue of the CanadianJournal of Surgery. The commentarydeals with the identification andmeas urement of operating room (OR)performance indicators, addressingthe variation among hospitals in termsof which indicators are collected andanalyzed.

Common definitions among hospi-tals are essential for external bench-marking. Although the authors identi-fied 8 indicators as the most criticalfor monitoring OR performance in15 children’s hospitals in Canada, def-initions for these indicators vary in lit-erature and across hospitals.

In the Netherlands, OR depart-ments of all 8 university medical cen-tres (UMCs) established a nationwidebenchmarking collaboration in 2005that is still active today. The objectiveof the collaboration is to improve ORperformance by learning from eachother through exchanging best prac-tices. Each UMC provides recordsfor all performed surgical cases to acentral OR benchmark database. Thisextensive database, presently com-prising more than 1 million surgicalcase records, is used to calculate keyperformance indicators related to theutilization of OR capacity. The data-base is also used for multicentreresearch on OR scheduling topicsand OR efficiency.

At the start of this collaboration, aset of performance indicators, particu-larly from a utilization perspective,was identified. Next, data definitionsof time periods and methods of reg -istration, as well as definitions of performance indicators, were har mon -ized among all benchmarking par -ticipants, a process that took nearly2 years. An independent data manage-ment centre enters the longitudinaldata collection in the central ORbenchmark database. This centre pro-vides professional expertise by facili-tating and processing data, and by per-forming reliability checks before dataare deemed ready for analysis.

LETTERS TO THE EDITOR / COMMENTARIES

Lette

rs to

the

Edi

tor /

Com

men

tarie

s

Sam

enva

tting

A

ckno

wle

dgm

ents

D

ankw

oord

P

ublic

atio

ns a

nd P

rese

ntat

ions

257

Lette

rs to

the

Edi

tor /

Com

men

tarie

s

Sam

enva

tting

A

ckno

wle

dgm

ents

D

ankw

oord

P

ublic

atio

ns a

nd P

rese

ntat

ions

E104 J can chir, Vol. 56, No 5, octobre 2013

CORRESPONDANCE

Our collaboration frequently meetsto discuss data analysis results andexplore processes and practicesbeyond the data. Through promotingdialogue among UMCs, a learningenvironment has been created.

G. Kazemier, MD, PhDProfessor of Hepatobiliary Surgery and Transplantation

Department of SurgeryVU University Medical CenterAmsterdam, the NetherlandsE. van Veen-Berkx, MScErasmus University Medical CenterDepartment of Operating RoomsRotterdam, the Netherlandsfor the Dutch OR Benchmarking Collaboration


DOI: 10.1503/cjs.020813

THE AUTHORS RESPOND

We thank Dr. Stepaniak for his inter-est in our commentary on the identi-fication and use of operating room(OR) efficiency indicators. While theProcedural Times Glossary, develop -ed by the Association of AnesthesiaClinical Directors (AACD), is a lead-ing source of procedural time defini-tions in support of economic and effi-ciency analyses within the OR, thisdoes not negate the fact that variableperformance indicator definitionsnonetheless exist in the body of ORefficiency literature. Moreover, des -pite the availability of leading sourcesof definitions such as the ProceduralTimes Glossary, differences in howhospitals define key OR performanceindicators persist.

Furthermore, even the AACD’sProcedural Times Glossary may notalways be adequate if one wants to

ensure consistent performance indica-tor data collection across multiplehospitals. For example, the AACDdefines “turnover time” as the “timefrom prior patient out of room to suc-ceeding patient in room time forsequentially scheduled cases.”1 How-ever, while this definition is clearlymeant to exclude idle time betweennonsequentially scheduled cases, itdoes not entirely address potentialexclusions, such as delays betweensequentially scheduled cases unrelatedto room cleaning and preparation(e.g., patient arrives late); how thesesituations are handled varies signifi-cantly across hospitals and materiallyimpacts how the indicator is collected.

Another example is the definitionof “on-time starts,” defined as thepatient being in the OR at the sched-uled time.1 This does not consider,however, whether certain late startsshould be excluded (e.g., owing todelayed access to postoperative beds,as is the case at some hospitals).

Thus, we do believe that there isroom for professional associations toagree to develop common metrics andoperational definitions, perhaps usingthe AACD’s Procedural Times Glos-sary (or an equivalent source) as astarting point, closing any gaps fromthere.

Regarding Dr. Stepaniak’s secondpoint, while performance indicatorsmay not contribute to the effective useof resources as defined by the Instituteof Medicine’s Committee on QualityHealth Care in America, they may doso under another definition, such asthe Oxford English Dictionary, whichdefines “effective” as “having anintended or expected effect.” If usingresources efficiently leads to the mostpatients having surgery in the best way

(i.e., on time starts, no delays, no can-cellations), then use of OR perform -ance indicators to monitor operationalperformance can indeed lead to theeffective use of resources.

In addition, we also thank Dr.Kazemier and Ms. van Veen-Berkx fortheir interest in our commentary andnote that the Dutch experience,whereby it took 2 years to harmonizeOR performance indicator definitionsand reporting across 8 universitymedical centres, speaks to the com-plexity of the undertaking and thecontinuing lack of universal standardsfor indicator definitions.

Moreover, some Canadian prov -inces have also had some success inharmonizing OR performance indi-cators, such as the OR BenchmarksCollaborative in Ontario. As ourcommentary has demonstrated,though, variable indicator definitionspersist and harmonizing them nation-ally may be particularly challengingdue to the provincial delivery ofhealth care.

Tamas Fixler, MASc, MBAIBM Canada Ltd.Thornhill, Ont.James G. Wright, MD, MPHDepartment of SurgeryRobert B. Salter Chair of Pediatric Surgical Research

The Hospital for Sick ChildrenToronto, Ont.


DOI: 10.1503/cjs.020513

Reference

1. Donham RT. Defining measurable OR-PRscheduling, efficiency, and utilization dataelements: the Association of AnesthesiaClinical Directors’ procedural times glos-sary. Int Anesthesiol Clin 1998;36:15-29.

258258

SYSTEMS-LEVEL QUALITY IMPROVEMENT

Comment on Research Article Entitled “Variability

of Subspecialty-Specific Anesthesia-Controlled Times at Two

Academic Institutions” as published in J Med Syst 2014; 38 (11)

Elizabeth van Veen-Berkx & J. Bitter & S. G. Elkhuizen & W. F. Buhre &

C. J. Kalkman & H. G. Gooszen & G. Kazemier &

for the Dutch Operating Room Benchmarking Collaborative

Received: 6 February 2014 /Accepted: 31 March 2014# Springer Science+Business Media New York 2014

Dear Dr. Ehrenfeld,

With profound interest we read the article written by Kodali,

Kim, Flanagan, Urman and you in the February 2014 issue of

the Journal of Medical Systems [1]. The article dealt with a

large dataset retrieved from two American academic institu-

tions and analyzed anesthesia-controlled times (ACT) per

subspecialty service, thereafter compared them to previously

published ACT data. The authors concluded that individual

specialty-specific ACT should be used to improve operating

room (OR) scheduling and to benchmark anesthesia

performance.

We could not agree more with the content and conclusions

of this interesting and well-executed study. The publication

stated that little work has been done to establish ACT

benchmarks for heterogeneous tertiary care centers. This

is an accurate statement, however, we would like to

provide additional benchmark data concerning ACT. In

the Netherlands, OR departments of all eight University

Medical Centers (UMCs) established a nationwide

benchmarking collaborative in 2005, which is still active

today [2, 3, 4]. The objective is to improve OR perfor-

mance by learning from each other through exchanging

good practices. Each UMC provides data records for all

surgical cases performed to a central OR benchmark

database. This extensive database, presently comprising

more than one million records of surgical cases, is used

to calculate key performance indicators related to the

utilization of OR capacity.

This article is part of the Topical Collection on Systems-Level Quality

Improvement

Electronic supplementary material The online version of this article

(doi:10.1007/s10916-014-0051-z) contains supplementary material,

which is available to authorized users.

E. van Veen-Berkx (*)

Department of Operating Rooms, Erasmus University Medical

Center Rotterdam, Room number: Hs-324, 's-Gravendijkwal 230,

3015 CE Rotterdam, The Netherlands

e-mail: [email protected]

E. van Veen-Berkx

e-mail: [email protected]

J. Bitter

Department of Operating Rooms, Bernhoven Hospital,

Nistelrodeseweg 10, 5406 PT Uden, The Netherlands

S. G. Elkhuizen

Health Policy and Management, Erasmus University Rotterdam,

Burgemeester Oudlaan 50, 3062 PA Rotterdam, The Netherlands

W. F. Buhre

Department of Anesthesiology and Pain Medicine, University

Medical Center Maastricht, P. Debyelaan 25, 6229 HX Maastricht,

The Netherlands

C. J. Kalkman

Department of Anesthesiology, University Medical Center Utrecht,

Heidelberglaan 100, 3584 CX Utrecht, The Netherlands

H. G. Gooszen

Department of Operating Rooms, Radboud University Medical

Center, Geert Grooteplein-Zuid 10, 6525 GA Nijmegen,

The Netherlands

G. Kazemier

Department of Surgery, VU University Medical Center, Amsterdam,

Room number: ZH7F011, De Boelelaan 1117, 1081 HVAmsterdam,

The Netherlands

J Med Syst (2014) 38:51

DOI 10.1007/s10916-014-0051-z

Lette

rs to

the

Edi

tor /

Com

men

tarie

s

Sam

enva

tting

A

ckno

wle

dgm

ents

D

ankw

oord

P

ublic

atio

ns a

nd P

rese

ntat

ions

259

Lette

rs to

the

Edi

tor /

Com

men

tarie

s

Sam

enva

tting

A

ckno

wle

dgm

ents

D

ankw

oord

P

ublic

atio

ns a

nd P

rese

ntat

ions

The database is also used for multicenter research on OR

scheduling topics and OR efficiency. Recently, our manuscript

entitled “The Influence of Anesthesia-Controlled Time on

Operating Room Scheduling in Dutch University Medical

Centers [4]” was published in The Canadian Journal of

Anesthesia/Journal canadien d’anesthésie. This publication

also provides ACT benchmarks per surgical department

based on an extensive multicenter dataset (N=330,258;

6 UMC’s), see Tables 1 and 2 in this letter (Tables 2 and

3 of our article in The Canadian Journal of Anesthesia/

Journal canadien d’anesthésie). Correspondingly, we con-

clude that efficient OR scheduling demands the accurate

prediction of surgeon-controlled time (SCT) as well as

ACT. Based on our dataset, we advise grossing up the

SCT by 33 % to account for ACT, as opposed to

employing a fixed number of minutes methodology for

ACT, which is the common practice in many hospitals in

the Netherlands.

Yours sincerely

Appendix 1

Performance indicators Dutch Operating Room Benchmarking

Collaborative

References

1. Kodali BS, Kim KD, Flanagan H, Ehrenfeld JM, Urman RD (2014)

Variability of Subspecialty-Specific Anesthesia-Controlled Times at Two

Academic Institutions. JMed Syst 38:11. doi:10.1007/s10916-014-0011-7

2. Kazemier G, Van Veen-Berkx E (2013) Comment on Identification

and use of operating room efficiency indicators: the problem of defi-

nition. Can J Surg 56:E103–4. doi:10.1503/cjs.020813

3. Van Veen-Berkx E, Elkhuizen SG, Kalkman CJ, Buhre W F, and

Kazemier G (2013) Successful Interventions to Reduce First-Case

Tardiness in Dutch University Medical Centers: Results of a

Nationwide Operating Room Benchmark Study. Am J Surg. doi:10.

1016/j.amjsurg.2013.09.025.

4. Van Veen-Berkx E, Bitter J, Elkhuizen SG, Buhre WF, Kalkman CJ,

Gooszen HG, Kazemier G (2014) The Influence of Anesthesia-Controlled

Time onOperating Room Scheduling inDutchUniversityMedical Centers.

Can J Anesth/J Can Anesth doi:10.1007/s12630-014-0134-9

Table 1 (Table 2 in the CJA ar-

ticle) descriptive statistics of ac-

tual total procedure time, actual

anesthesia-controlled time and

actual surgeon-controlled time

(all in minutes), as registered in

the central OR benchmark

database

UMC N Total procedure time Anesthesia-Controlled time Surgeon-Controlled time

Mean SD Median Mean SD Median Mean SD Median

UMC1 34,316 160 109 131 34 18 30 126 101 100

UMC2 52,329 181 126 142 43 24 36 138 112 104

UMC3 70,264 178 123 146 44 24 39 134 110 104

UMC4 41,266 152 121 113 32 17 27 120 112 84

UMC5 45,955 162 120 130 36 20 31 126 108 96

UMC6 86,128 127 104 92 30 20 26 97 92 65

Total 330,258 158 119 124 37 22 31 121 106 90

Table 2 (Table 3 in the CJA article) descriptive statistics of actual total procedure time, actual anesthesia-controlled time and actual surgeon-controlled

time (all in minutes), as registered in the central OR benchmark database, differentiated per surgical department using the data of all six UMCs

Actual total procedure time Anesthesia-Controlled time Surgeon-Controlled time

N Mean SD Median Mean SD Median Mean SD Median

Cardiothoracic surgery 29,408 264 115 261 59 25 56 205 106 201

General surgery 76,203 173 120 143 40 24 34 133 106 106

Ear-Nose-Throat surgery 41,551 129 113 93 31 16 29 98 105 61

Oral & Maxillofacial surgery 13,170 165 130 130 38 18 35 127 121 94

Neurosurgery 23,969 216 143 170 45 24 40 171 132 128

Ophthalmology 36,086 77 41 69 21 12 19 56 35 49

Orthopedic surgery 35,184 148 86 134 35 20 31 112 77 100

Plastic surgery 24,001 148 127 112 32 19 28 116 118 82

Urology 27,210 134 101 99 32 17 28 102 92 70

Obstetrics & Gynaecology 23,476 138 92 113 33 17 29 105 83 82

Total 330,258 158 119 124 37 22 31 121 106 90

51, Page 2 of 2 J Med Syst (2014) 38:51

260

SAMENVATTING

De centrale vraag van dit proefschrift luidt als volgt: leidt een nationaal, lange termijn, samenwerkingsverband van de operatiekamerafdelingen (OK-afdelingen) van alle acht de universitair medische centra (UMCs) in Nederland tot verbeteringen in de overall bedrijfsvoering van operatiekamers (OKs)? Om deze vraag te beantwoorden, zijn meerdere, bijna allemaal multicenter, studies uitgevoerd:

één exploratief, multicenter onderzoek waarin kwalitatieve en kwantitatieve methoden zijn gecombineerd;

drie descriptieve, kwantitatieve, multicenter studies gebaseerd op een substantiële hoeveelheid OK-data;

en zes quasi-experimentele, kwantitatieve, meestal multicenter studies die het effect van specifieke interventies in diverse OK-processen onderzoeken.

Hoofdstuk 1: Het Benchmarken van OK-afdelingen in NederlandHet doel van deze studie was om te onderzoeken of het lange termijn samenwerkingsverband Benchmarking OK heeft geleid tot voordelen voor het OK-management en voor de overall bedrijfsvoering van OKs. Hiertoe is het samenwerkingsverband geëvalueerd aan de hand van een reeds ontwikkeld evaluatiekader voor benchmarkactiviteiten en met behulp van ‘mixed methods research’ waarbij kwalitatieve en kwantitatieve onderzoeksmethoden zijn gecombineerd.

Benchmarking op een collaboratieve manier, zoals de acht Nederlandse UMCs doen met Benchmarking OK, is een vruchtbare methode als het gaat om het identificeren van verbeterpotentieel. Ook wordt het bij de UMCs ingezet als een continu proces ter verbetering van de eigen OK-organisatie. Dit onderzoek toonde aan dat benchmarking hiernaast nog meer voordelen heeft. Het is noemenswaardig dat Benchmarking OK, gestart in 2004, nog steeds aanhoudt terwijl het al meer dan tien jaar bestaat. “Het doel van het netwerken” werd door alle respondenten genoemd als het belangrijkste voordeel. De netwerkbijeenkomsten (kerngroepvergaderingen, studiemiddagen en congressen) die werden georganiseerd door Benchmarking OK maakten het eenvoudiger voor deelnemers om ook nog eens één-op-één contact op te nemen met een collega werkzaam in een ander UMC of om bij elkaar op bezoek te gaan. Dergelijke informele, onderlinge contacten hebben bijgedragen aan het delen en verspreiden van kennis en beleidsdocumenten (‘mooie voorbeelden’), en het initieerde verbeteractiviteiten op de OK. Juist het fysiek bij elkaar brengen van collega’s uit de verschillende UMCs was van belang om de niet-vastgelegde en niet-tastbare componenten van mooie voorbeelden en elders behaalde verbeteringen met elkaar te delen en te bediscussiëren. Dergelijke informatie is moeilijk te delen in meer formeel vastgelegde

261

Lette

rs to

the

Edi

tor /

Com

men

tarie

s

Sam

enva

tting

A

ckno

wle

dgm

ents

D

ankw

oord

P

ublic

atio

ns a

nd P

rese

ntat

ions

communicatiemiddelen zoals beleidsdocumenten of mailverkeer. In de bijeenkomsten werd ook niet alleen gesproken over de successen maar juist ook de knelpunten en ervaringen met methoden die niet werken op een OK. Juist het bespreken van ook de faalfactoren draagt bij aan het leren van elkaar.

Benchmarking is een continu proces dat door de UMCs wordt gebruikt om van elkaar te leren en ter verbetering van bedrijfsvoering van de eigen OK-organisatie. Ook al heeft Benchmarking OK, zoals benchmarkactiviteiten over het algemeen hebben, het doel om prestaties te verbeteren, daadwerkelijke, aantoonbare en meetbare verbeteringen zijn niet noodzakelijk om het samenwerkingsverband te continueren. De relatief beperkte focus op OK-benutting in deze benchmark blijkt een startpunt te zijn voor het uitwisselen van een hoeveelheid aan informatie en ervaringen betreffende de structuur, processen en prestaties van operatiekamerafdelingen, zowel op het gebied van bedrijfsvoering, kwaliteit en veiligheid. Nader onderzoek zou zich moeten richten op de relatie tussen benchmarking als instrument en de daadwerkelijke prestatieverbeteringen gerealiseerd in de lokale UMCs als gevolg van benchmarkactiviteiten.

Hoofdstuk 2: Verbetermogelijkheden in OK-benuttingHet doel van deze studie was het bepalen van mogelijkheden om het benutten van OK-tijd te verbeteren. Daartoe zijn de directe en indirecte relaties tussen late start, wisseltijd, vroeg-einde-leegstand en netto benutting onderzocht, en is bepaald welke prestatieindicator de meest negatieve impact heeft op de OK-benutting. Hiernaast zijn de indirecte relaties tussen de drie indicatoren die de leegstand van een OK meten (late start, wisseltijd en vroeg-einde-leegstand) onderzocht, om een zogenaamd ‘trickle down’-effect te identificeren naarmate de OK-dag vordert.

Meervoudige lineaire regressie en mediatie-effect-analyse zijn toegepast op een dataset van alle acht de UMCs in Nederland. Deze dataset bestond uit 190.071 OK-dagen. Op die dagen zijn 623.871 operaties uitgevoerd.

Vroeg-einde-leegstand aan het einde van de dag had de grootste negatieve invloed op OK-benutting, gevolgd door late start en daarna wisseltijd. De relaties tussen de drie indicatoren die de leegstand meten waren verwaarloosbaar. De impact van de partiële, indirecte effecten van die specifieke indicatoren op netto benutting waren statistisch significant, echter relatief klein. Het trickle-down-effect dat te laat starten aan het begin van de OK-dag meer vertraging kan veroorzaken gedurende het verloop van die OK-dag, wordt niet bevestigd door deze onderzoeksresultaten. Dit suggereert dat een vertraging opgelopen aan het begin van de dag wordt ingehaald.

Het verminderen van vroeg-einde-leegstand aan het einde van de dag kan de OK-benutting verbeteren. Het verbeteren van OK-planning, specifiek de voorspelbaarheid van de totale operatieduur, door het wijzigen van de volgorde van operaties, het aanpassen van

262

het beleid rondom het afzeggen van operaties, alsmede het flexibeler inroosteren van OK-medewerkers in overeenstemming met vraag en aanbod van patiënten, zijn manieren om vroeg-einde-leegstand te verminderen.

Hoofdstuk 3: De invloed van anesthesietijd op OK-planning in Nederlandse UMCsHet voorspellen van de totale operatieduur omvat meerdere onderdelen die onderhevig zijn aan variabiliteit, inclusief de twee hoofdcomponenten: chirurgische tijd en anesthesietijd (inleidingstijd en uitleidingstijd). In deze studie wordt onderzocht wat het effect op de OK-planning is als anesthesietijd wordt gezien als een proportioneel toenemend onderdeel van de totale operatieduur in plaats van een standaard aantal minuten per ingreep. Doelstelling is het verbeteren van de voorspelbaarheid van de totale operatieduur en daarmee het verbeteren van de overall OK-planning.

Data van zes UMCs en zeven opeenvolgende jaren (2005-2011) zijn geïncludeerd: in totaal 330.258 klinische, electieve patiënten. Het plannen ofwel “voorspellen” van de totale operatieduur, inclusief anesthesietijd, is herzien en in theorie opnieuw bepaald als de gerealiseerde chirurgische tijd * 1.33. Verschillen tussen de gerealiseerde en geplande operatieduur zijn bepaald voor beide planningsmethoden.

De theoretische planningsmethode, waarbij anesthesietijd wordt bepaald als een proportie van de chirurgische tijd, liet een verbeterde voorspelbaarheid van de totale operatieduur zien.

Efficiënt OK-management vraagt om een nauwkeurige voorspelling van de tijd die nodig is voor alle onderdelen van een operatie, inclusief chirurgische tijd en anesthesietijd. Op basis van een extensieve dataset van zes UMCs, is het advies om ten behoeve van de OK-planning 33% van de chirurgische tijd op te tellen bij de reeds geplande chirurgische tijd, om de planning van anesthesietijd te dekken. Dit in tegenstelling tot de huidige en meest gebruikte methode om voor anesthesietijd een standaard aantal minuten (meestal 20 minuten) in te plannen ongeacht de soort chirurgische ingreep en ongeacht de anesthesietechniek. Dit advies zal de OK-planning overall verbeteren en dit kan resulteren in een reductie van uitloop en afgezegde operaties aan het einde van de dag. Zodoende zal er efficiënter gebruik gemaakt kunnen worden van de beperkte en dure OK-capaciteit.

Hoofdstuk 4: Het effect van de individuele chirurg en anesthesioloog op OK-tijdVariatie in operatieduren veroorzaakt uitloop en leegstand van beschikbare OKs en OK-tijd. Eerder onderzoek heeft aangetoond dat voor een bepaalde procedure de chirurg de voornaamste bron van variatie is. Op dit moment is er geen berekening beschikbaar over de variatie in operatieduur veroorzaakt door de chirurg en de exacte effecten van individuele chirurgen op het voorspellen van de operatieduur. Daarom is dat het doel van deze studie.

263

Lette

rs to

the

Edi

tor /

Com

men

tarie

s

Sam

enva

tting

A

ckno

wle

dgm

ents

D

ankw

oord

P

ublic

atio

ns a

nd P

rese

ntat

ions

Omdat anesthesietijd een belangrijk onderdeel van de totale operatieduur vormt, wordt in deze studie ook de variatie tussen anesthesiologen gekwantificeerd. De analyse betreft 16,480 patiënten van de afdeling algemene heelkunde. De totale hoeveelheid aan variatie in de operatieduur veroorzaakt door type procedure, eerste operateur, tweede operateur en anesthesioloog was bepaald met behulp van linearmixedmodels. Het effect daarvan op het voorspellen (plannen) van de operatieduur werd geëvalueerd aan de hand van de duur van de uitloop en de duur van de leegstand aan het einde van de dag.

De verschillen tussen eerste operateurs verklaren slechts 2,9% en de verschillen tussen anesthesiologen slechts 0,1% van de variatie in OK-tijd. Rekening houden met individuele chirurgen en anesthesiologen in de OK-planning leidt tot een gemiddelde reductie van 1,8 minuten uitloop en 3,0 minuten leegstand. In vergelijking met type procedure (nog steeds de voornaamste oorzaak voor variatie) kunnen de verschillen tussen chirurgen maar een klein deel van de variatie in OK-tijd verklaren. De impact van de verschillen tussen anesthesiologen op OK-tijd is te verwaarlozen. Een voorspellingsmodel dat rekening houdt met de verschillen individuele chirurgen en anesthesiologen kan de precisie doen toenemen, echter verbeteringen zijn te marginaal om consequenties te hebben voor de OK-planning in de dagelijkse praktijk.

Hoofdstuk 5: Interventies om late start te verminderen in Nederlandse UMCs. Resultaten van een landelijke OK-benchmarkstudie“Late start” ofwel het te laat starten van de eerste operatie die gepland staat op het OK-programma van een specifieke OK op een bepaalde OK-dag, is nog steeds een veel voorkomende frustratie. In deze studie is gebruik gemaakt van een landelijke OK-benchmark database van de acht UMCs in Nederland, om de effectiviteit van verschillende interventies te onderzoeken die geïmplementeerd zijn om late start te verminderen. Ook is de economische impact van deze interventies onderzocht.

OK-data van alle acht de Nederlandse UMCs en van zeven achtereenvolgende jaren is geïncludeerd: in totaal 190.295 klinische, electieve, eerste patiënten (per OK en per OK-dag). Data is geanalyseerd met SPSS Statistics en multidisciplinaire focusgroepen, waar operatieassistenten, anesthesiemedewerkers, anesthesiologen, chirurgen, OK-managers en interne organisatieadviseurs werkzaam in het ziekenhuis specifiek voor de OK, aan hebben deelgenomen. Alle acht de UMCs waren tijdens deze focusgroepen vertegenwoordigd door verschillende beroepsgroepen. ANOVA met contrasten is gebruikt als analysetechniek om de invloed van de interventies te bepalen.

De tijd die jaarlijks “verdwijnt” binnen de UMCs door het te laat starten van de eerste patiënt op het OK-programma heeft een aanzienlijke economische impact. In die verloren tijd hadden anders ook 9.707 operaties van 60 minuten (gemiddeld 173 operaties per UMC per jaar) uitgevoerd kunnen worden. Vier centra hebben interventies geïmplementeerd en daarmee een significante reductie in de vertraging aan het begin van de dag weten te

264

bewerkstelligen. Voorbeelden van succesvolle interventies zijn: het regelmatig informeren van de OK-teams over het te laat starten en directe feedback richting het OK-team wanneer te laat wordt gestart; afspraken met de Intensive Care over het creëren van een tijdelijk “overbed” indien er in de ochtend niet direct IC-capaciteit beschikbaar is, zodat de OK wel kan opstarten; niet de anesthesiemedewerker maar de operatieassistent wordt verantwoordelijk gemaakt voor het transport van de patiënt van de holding naar de OK, zodat de anesthesiemedewerker door kan gaan met het voorbereiden van de inleiding op de OK.

Deze studie toont aan dat het landelijk benchmarken van OK-data ingezet kan worden ten behoeve van het identificeren en meten van de effectiviteit van interventies om late start te verminderen in een academische ziekenhuissetting. De in vier centra geïmplementeerde interventies waren succesvol in het significant verminderen van vertraging aan het begin van de OK-dag.

Hoofdstuk 6: Reëel plannen van anesthesietijd vermindert afgezegde operaties en versoepelt de OK-workflow in een academische ziekenhuissettingOp de klinische OK-afdeling van het Academisch Medisch Centrum (AMC) Amsterdam is op 1 juli 2012 een nieuwe planningsmethode specifiek voor anesthesietijd (inleidingstijd en uitleidingstijd) geïmplementeerd. Deze studie onderzoekt de relatie tussen deze nieuwe planningsmethodiek en de OK-prestaties. De nieuwe methodiek omvatte het ontwikkelen van zogenoemde “anesthesiepakketten”. Aan ieder anesthesiepakket is een anesthesietechniek en de daarbij horende benodigde tijdsduur gekoppeld. Er bestaan zeven pakketten (0 tot en met 6), waaronder ook specifieke pakketten voor kinderen. Met behulp van deze pakketten kan de benodigde tijd voor anesthesie (inleidingstijd en uitleidingstijd) specifieker worden ingepland en afgestemd op de techniek die gebruikt gaat worden, de hoeveelheid monitoren en de complexiteit van de patiënt. Het anesthesiepakket dat gebruikt gaat worden tijdens de OK wordt bepaald en ingepland door de anesthesioloog tijdens het preoperatieve assessment op de poli, ruim voorafgaand aan de OK.

Een quasi-experimenteel tijdserie design is toegepast. De relevante data is verdeeld in vier gelijke tijdsperioden. Die perioden zijn onderling vergeleken door middel van een ANOVA met contrasten: een interventie-, pre-interventie- en post-interventie-contrast zijn geanalyseerd. Alle spoedpatiënten zijn geëxcludeerd. In totaal zijn 34.976 klinisch, electieve patiënten, uitgevoerd gedurende de periode van 1 januari 2010 tot en met 31 december 2014, geïncludeerd voor analyse.

De interventie-contrast toonde een significante 4,5% afname van de planningsafwijking ofwel de fout in de voorspelling. Het totaal aantal afgezegde operaties nam af met 19,9%. De ANOVA met contrasten toonde geen significante verschillen in het aantal minuten (en de frequentie) vroeg-einde-leegstand en uitloop aan het einde van de dag en ook geen verschillen in het netto benuttingspercentage. Tijdens deze studie kwamen meerdere

265

Lette

rs to

the

Edi

tor /

Com

men

tarie

s

Sam

enva

tting

A

ckno

wle

dgm

ents

D

ankw

oord

P

ublic

atio

ns a

nd P

rese

ntat

ions

positieve neveneffecten aan het licht die bijdragen aan een soepele workflow op de OK-afdeling, bijvoorbeeld: voorbeelden zijn dat anesthesiemedewerkers nu vooraf weten welk anesthesiepakket is ingepland en dus welke materialen voor de inleiding klaargelegd moeten worden en welke apparatuur van tevoren getest moet worden op gebruik.

Het reëler inplannen van de twee hoofdcomponenten van een operatie, chirurgische tijd en anesthesietijd, leidt tot minder afgezegde operaties, minder planningsafwijking en een meer soepele OK-workflow in een academische ziekenhuissetting.

Hoofdstukken 7, 8 en 9: Multidisciplinair teamwork verbetert de prestaties van de OKDe hoofdstukken 7, 8 en 9 beschrijven drie onderzoeken die het concept “WPM”: multidisciplinair werkplekmanagement op de OK en het effect van deze samenwerkingsvorm op de bedrijfsvoering van de OK. Een single-center, kwalitatieve, casestudie (hoofdstuk 7); een single-center, kwantitatief, longitudinaal onderzoek (hoofdstuk 8) en een multicenter studie met een quasi-experimenteel tijdserie design (hoofdstuk 9). Het concept WPM is ontwikkeld en geïmplementeerd in het Radboudumc te Nijmegen. De drie studies bevestigen het belang van multidisciplinaire, op team-gebaseerde zorg en multidisciplinaire samenwerking tussen professionals in de operatieve zorgketen. Goed functionerende WPM’s waren in staat om knelpunten in een vroeg stadium te signaleren en op te lossen, alsmede de continuïteit in de zorgketen te waarborgen (hoofdstuk 7). WPM’s verkrijgen met behulp van verschillende prestatie-indicatoren, zoals benuttingspercentages, uitloop en afgezegde operaties, inzicht in hun performance en hiermee ook de mogelijkheid om die performance te verbeteren.

Nader onderzoek (hoofdstuk 8) toont aan dat twee snijdende afdelingen in het Radboudumc door middel van het WPM gedurende de jaren (2005 – 2011) een geleidelijke verbetering in het netto benuttingspercentage weten te bewerkstelligen. Ieder jaar is er een significante daling zichtbaar van de variatie (interkwartielafstand Q3 – Q1) in netto benutting en een significante toename in de gemiddelde netto benutting sinds de implementatie van het WPM. De stapsgewijze daling van de variatie in het benuttingspercentage toont een organisatorisch leereffect aan, alsmede meer stabiliteit en voorspelbaarheid in het OK-planningsproces (het meer ‘in control’ zijn op dat vlak). De toename van het benuttingspercentage en de afname van onzekerheden in de OK-planning zijn kenmerken van een efficiënter gebruik van schaarse, kostbare OK-tijd. Bovendien versterkt het multicenter onderzoek (hoofdstuk 9) de gedachte dat multidisciplinaire samenwerking in de vorm van WPM’s ingezet in de perioperatieve keten, een positieve invloed heeft op OK-planning en de benutting van OK-tijd: gedurende de periode 2005 tot en met 2013 heeft Radboudumc namelijk de hoogste mediaan netto benutting, te weten 94% ten opzichte van 85%, de mediaan in de controlegroep (zes UMCs). Een bijkomend interessant detail is dat andere nationale databases met mortaliteitscijfers het idee ondersteunen dat WPM’s ook van belang kunnen zijn voor het verbeteren van de kwaliteit

266

en veiligheid van de zorg, omdat het Radboudumc, voor één specifiek snijdend specialisme, de laagste mortaliteits- en complicatiecijfers laat zien in vergelijking met datzelfde specialisme in de andere UMCs.

Hoofdstuk 10: Een dedicated spoed-OK gener eert een hogere benutting, minder uitloop en minder afgezegde operatiesOver het algemeen bestaan er twee methoden om om te gaan met spoed: (1) het standaard leeghouden van één (of soms zelfs twee) operatiekamers waarop spoedpatiënten worden opgevangen en (2) het vrijhouden van een beetje capaciteit in de planning van iedere operatiekamer, de ‘witte vlek’, om spoedpatiënten in op te vangen. Eerdere studies tonen tegenstrijdige resultaten met betrekking tot wat de beste methode zou zijn in relatie tot benutting van OK-tijd. Deze studie analyseert de empirische data van drie UMCs met behulp van een quasi-experimenteel, gecontroleerd, tijdserie design. Vier verschillende tijdsperioden zijn onderling vergeleken door middel van een ANOVA met contrasten.

De resultaten zijn gebaseerd op 467.522 patiënten in totaal. Na het sluiten van de spoed-OK steeg het netto benuttingspercentage enigszins; de duur van de uitloop echter ook. Dit is in tegenstelling tot de resultaten van de eerder uitgevoerde, theoretische, simulatiestudie, waarin methode 2 was gemodelleerd. De twee controle UMCs, die wel een spoed-OK behouden, vertoonden een hogere toename van het netto benuttingspercentage alsmede een afname van de duur van de uitloop én een lager percentage afgezegde operaties wegens spoed. Deze studie toont aan dat in de dagelijkse praktijk een dedicated spoed-OK de te prefereren methode is in relatie tot de OK-prestaties netto benutting, uitloop en afgezegde operaties. De resultaten onderschrijven ook de gedachte dat simulatieonderzoek, waarbij een deel van de zorgpraktijk in een wiskundig model wordt gegoten en gerepresenteerd in een computerprogramma, van belang is voor het onderbouwen van evidence-based beleidskeuzes. Echter, simulatieonderzoek is niet in staat om de volledige complexiteit van professionele organisaties, zoals operatiekamerafdelingen van (academische) ziekenhuizen en de zorgverleners die daarbinnen samenwerken, te vatten. Met andere woorden, de praktijk is weerbarstig.

267

Lette

rs to

the

Edi

tor /

Com

men

tarie

s

Sam

enva

tting

A

ckno

wle

dgm

ents

D

ankw

oord

P

ublic

atio

ns a

nd P

rese

ntat

ions

268

Acknowledgments

The authors sincerely thank all members of the steering and project committee of the Dutch Operating Room Benchmarking Collaborative for their participation and contribution.

Current Members of the Dutch Operating Room Benchmarking Collaborative (the Netherlands):

Academic Medical Center Amsterdam: Ron Balm, MD, PhD, Department of Vascular Surgery; Vincent M. de Jong, MSc, Department of Surgery; Diederich C.C. Cornelisse, MSc, Department of Operating Rooms;

MaastrichtUniversityMedicalCenter:Wolfgang F. Buhre, MD, PhD, Division of Anesthesiology; Hub J. Ackermans, Department of Operating Rooms;

ErasmusUniversityMedicalCenterRotterdam: Robert Jan Stolker, MD, PhD, Department of Anesthesiology and Division of Emergency, Perioperative and Intensive Care; Jeanne Bezstarosti, MD, Department of Anesthesiology and Department of Operating Rooms;

LeidenUniversityMedicalCenter:Roald R. Schaad, MD, Department of Anesthesiology; Jos H. Wind, Department of Operating Rooms;

University Medical Center Groningen: Irmgard Krooneman-Smits, MBA, Department of Operating Rooms; Peter Meyer, MD, PhD, Department of Anesthesiology and Department of Operating Rooms;

RadboudUniversityMedicalCenterNijmegen: Simon A.W. Broecheler, MSc, Department of Operating Rooms and Department of Anesthesiology; Mirjam van Dijk-Jager, Department of Operating Rooms;

UniversityMedicalCenterUtrecht:A. Christiaan Kroese, MD, Department of Anesthesiology and Department of Operating Rooms; Jeffrey Kanters, Department of Operating Rooms;

UniversityofTwente: Johannes J. Krabbendam, PhD; Erwin W. Hans, PhD, Department of Operational Methods for Production and Logistics;

269

Lette

rs to

the

Edi

tor /

Com

men

tarie

s

Sam

enva

tting

A

ckno

wle

dgm

ents

D

ankw

oord

P

ublic

atio

ns a

nd P

rese

ntat

ions

VU University Medical Center: Derk P. Veerman, MD, PhD, Division of Anesthesiology, Intensive Care and Emergency Medicine; Kjeld H. Aij, PhD, MBA, Division of Anesthesiology, Intensive Care and Emergency Medicine. Lizan Aalders, MSc, Department of Operating Rooms.

The authors also sincerely thank Ewout W. Steyerberg, PhD, Professor of Medical Decision Making, and Daan Nieboer, Researcher, Department of Public Health, Erasmus University Medical Center Rotterdam, for their statistical advice.

270

DANKWOORD

Dit proefschrift heb ik niet alleen geschreven, vele mensen hebben mij tijdens het onderzoek geholpen en gesteund. Ik wil hen allen hartelijk danken. Een paar mensen wil ik in het bijzonder bedanken:

Professor Geert Kazemier, promotor, zonder jou geen Benchmarking OK en zonder jou geen proefschrift. Wat ben ik blij dat je mij hebt laten zien hoe leuk het is om onderzoek te doen en diepgang te vinden in de verhalen achter de OK-cijfers. De start kende zeker meerdere hobbels maar dankzij jouw intensieve begeleiding tijdens het schrijven van het eerste artikel, kreeg ik de smaak te pakken. Wat heb ik veel van je geleerd. Ik hoop dat we in de toekomst ook nog eens samen een artikel zullen schrijven. Veel dank voor je vertrouwen.

Professor Hein Gooszen, tweede promotor. Dank voor de wetenschappelijke begeleiding en voor het altijd gastvrije onthaal in Nijmegen. Wat zijn het toch fijne mensen om mee samen te mogen werken op de OK in het Radboudumc.

Professoren Cor Kalkman en Wolfgang Buhre, dank voor jullie kritische revisies en het sparren over het merendeel van de publicaties in dit proefschrift.

Sylvia Elkhuizen, jouw statistische en methodologische ondersteuning was altijd zeer welkom en ook hard nodig. Veel dank hiervoor.

Masterstudenten Health Care Management aan het instituut Beleid, Management & Gezondheidszorg van de Erasmus Universiteit Rotterdam: Bart, Sanne en Olivier in het bijzonder. Wat een voorrecht om jullie te mogen begeleiden tijdens jullie afstudeerscripties. Niet alleen voor jullie maar ook voor mij was dat erg leerzaam.

De kerngroep Benchmarking OK: Jeanne Bezstarosti, Diederich Cornelisse, Lizan Aalders, Hub Ackermans, Jos Wind, Peter Meyer, Jeffrey Kanters, Mirjam van Dijk-Jager en Jeffrey Kanters. De stuurgroep Benchmarking OK: Robert Jan Stolker, Ron Balm, Wolfgang Buhre, Roald Schaad, Irmgard Krooneman-Smits, Simon Broecheler, Christiaan Kroese, Peter Veerman en Kjeld Aij. Hartelijk dank voor de altijd fijne samenwerking en de gastvrijheid. En dank voor jullie openheid over het reilen en zeilen op de OK-afdelingen van de UMCs. Wat is het toch een bijzondere en ook complexe ‘wereld’ in het ziekenhuis. Ik hoop dat we elkaar nog regelmatig tegen zullen komen.

271

Lette

rs to

the

Edi

tor /

Com

men

tarie

s

Sam

enva

tting

A

ckno

wle

dgm

ents

D

ankw

oord

P

ublic

atio

ns a

nd P

rese

ntat

ions

Justin Bitter, professor Gert Jan Scheffer, professor Pierre van Amelsvoort, Fleur Mokken, Menno van Dijk, Diederich Cornelisse, Dirk de Korne, professor Roland Bal, Ruben van Eijk en professor Marinus Eijkemans: bedankt voor jullie ideeën en het co-auteurschap.

Beste paranimfen, Aeltke Dekker en Marleen Moerman, beter bekend als Marleen van Dijkman. Dank voor jullie hulp bij de totstandkoming van deze dag en wat fijn dat jullie letterlijk ‘naast’ mij staan vandaag. Aeltke, we hebben elkaar leren kennen in de collegebanken van de EUR en daar ben ik nog steeds erg blij om. Wat ben ik trots op je dat je nu weer in diezelfde collegebanken zit om Gezondheidsrecht te studeren. Marleen, carrièrevrouw, het voorbeeld van doorzettingsvermogen. Fijn dat ik altijd even met je kon sparren over het onderzoek en het vele schrijfwerk, ook als ik het eventjes niet meer zag zitten.

Nick, bedankt voor de interesse die je altijd hebt in mijn werk en het onderzoek. Laten we nooit stoppen met onze, soms stev ige, discussies over de gezondheidszorg. Lieve Editha, ik mis hoe je was en zal dat nooit vergeten. Ik had graag gewild dat ik je langer helemaal gezond had meegemaakt. Naomi, Bart-Jan, Raoul en Mireille, wat fijn om een grote schoonfamilie in de buurt te hebben en bedankt dat de kindjes regelmatig bij jullie in een vertrouwde en altijd gezellige omgeving terecht kunnen als wij moeten werken.

Lieve papa en mama, bedankt voor jullie steun en altijd goede zorgen. Bij jullie kan ik altijd terecht voor een luisterend oor en voor bemoedigende woorden. Jullie goede zorgen komen vaak in de vorm van lekkere Aziatische gerechten en heerlijke baksels, dank je wel, mama! Papa, regelmatig ontvang ik van jou actuele stukken over de gezondheidszorg maar ook over opvoeding, het gezinsleven en ‘werkende moeders’, de onderwerpen die voor mij van belang zijn. Blijf dat vooral doen, want jij zorgt ervoor dat ik kritisch blijf en dat ik onderwerpen vanuit verschillende perspectieven blijf bekijken. Lieve papa en mama, ik hou van jullie.

Lieve broer, Matthieu, ook bij jou kan ik altijd terecht. En wat is het genieten om jou samen met Mees te zien! Jouw gedrevenheid en arbeidsethos zijn een voorbeeld voor mij: workhard,playhard. Ik hou van je, en ook van Wes.

Lieve Ruben, een tijd geleden grapte je nog: “als je straks gepromoveerd bent, kunnen we eindelijk weer eens op gelijk niveau met elkaar praten”, nu is het dan zover! Gelukkig hebben we hetzelfde gevoel voor humor. Bedankt voor je steun en motiverende woorden voorafgaand en tijdens dit promotietraject en zeker ook voor de welkome afleiding als ik wilde doorwerken. Lieve Rub, wat prijs ik mezelf gelukkig met jou en met Mees & Lieve. Wat een prachtgezin hebben wij samen gemaakt. Het duurde even, maar dan heb je ook wat! Ik hou met heel mijn hart van jullie.

272

LIST OF PUBLICATIONS

Van Veen-Berkx E, Kazemier G. Comment on “Identification and use of operating room efficiency indicators: the problem of definition”. Can J Surg 2013;56:E103-4.

Bitter J, Van Veen-Berkx E, Gooszen HG, van Amelsvoort P. Multidisciplinary teamwork is an important issue to healthcare professionals. Team Performance Management 2013;19:263-78.

Van Veen-Berkx E, Elkhuizen SG, Kalkman CJ, Buhre WF, Kazemier G. Successful interventions to reduce first-case tardiness in Dutch university medical centers: Results of a nationwide operating room benchmark study. Am J Surg 2014;207:949-59.

Van Veen-Berkx E, Bitter J, Elkhuizen SG, Buhre WF, Kalkman CJ, Gooszen HG, Kazemier G. The influence of anesthesia-controlled time on operating room scheduling in Dutch university medical centres. L’influence du temps controle par l’anesthesie sur le programme operatoire dans les centres medicaux d’une universite neerlandaise. Can J Anaesth 2014;61:524-32.

Van Veen-Berkx E, Bitter J, Elkhuizen SG, Buhre WF, Kalkman CJ, Gooszen HG, Kazemier G. Comment on Research Article Entitled “Variability of Subspecialty-Specific Anesthesia-Controlled Times at Two Academic Institutions” as published in J Med Syst 2014; 38(11). J Med Syst 2014;38:51.

Bitter J, Van Veen-Berkx E, van Amelsvoort P, Gooszen HG. Preoperative cross functional teams improve operating room performance. J Health Organ Manag 2015;29(3):343-52.

Van Veen-Berkx E, Bitter J, Kazemier G, Scheffer GJ, Gooszen HG. Multidisciplinary teamwork improves utilization of the operating room: a multicenter study. J Am Coll Surg 2015;220:1070-6.

Van Veen-Berkx E¸ Elkhuizen SG, van Logten S, Buhre WF, Kalkman CJ, Gooszen HG, Kazemier G. Enhancement opportunities in operating room utilization; with a statistical appendix. J Surg Res 2015;194:43-51.

Van Veen-Berkx E, Kuijper B, Elkhuizen SG, Kazemier G. Dedicated operating room for emergency surgery generates more utilization and less overtime. Am J Surg 2016;211:122-8.

Van Veen-Berkx E, de Korne DF, Olivier OS, Bal RA, Kazemier G. Hospital Benchmarking: Evaluation of an initiative between OR departments of university medical centers in the Netherlands. Benchmarking: An International Journal 2016;23(5):1171-92

Van Eijk RP, Van Veen-Berkx E, Kazemier G, Eijkemans MJ. Effect of Individual Surgeons and Anesthesiologists on Operating Room Time. Anesth Analg 2016;123:445-51.

Van Veen-Berkx E, Van Dijk MV, Cornelisse DC, Kazemier G, Mokken FC. Scheduling Anesthesia-Controlled Time Reduces Case Cancellations and Improves Operating Room Workflow. J Am Coll Surg 2016;223:343-51.

273

Lette

rs to

the

Edi

tor /

Com

men

tarie

s

Sam

enva

tting

A

ckno

wle

dgm

ents

D

ankw

oord

P

ublic

atio

ns a

nd P

rese

ntat

ions

PUBLICATIONS IN DUTCH

Van Veen-Berkx E, Schoch SLW. Concentratie van zorg vergt ommekeer in de OK. OK Operationeel, Managementkatern. 2014:28-30.

Van Veen-Berkx E, Kazemier G, Schoch SLW. Bedrijfsvoering OK-complexen gebaat bij benchmarking. OK Operationeel, Managementkatern. 2012:32-33.

Van Veen-Berkx E, Kazemier G. Benchmarking maakt prestaties op de OK transparant. Nederlands Tijdschrift voor Heelkunde.2012:177-180.

PRESENTATIONS AND EDUCATIONAL ACTIVITIES

2011-2015 Guest Lectures on Benchmarking and Operating Room Management, Master of Health Care Management at the Institute of Health Policy and Management, Erasmus University Rotterdam and University of Twente.

2012 & 2014 Scientific poster presentations on Operating Room Scheduling at the International Forum on Quality & Safety in Healthcare, Paris, France.

2011-2014 Presentations and workshops on Benchmarking and Operating Room Management at national conferences in the Netherlands.

DISSERTATION COMMITTEEPr omotorenProf.dr. G. KazemierProf.dr. H.G. Gooszen

LeescommissieProf.dr. H.J. BonjerProf.dr. S.A. LoerProf.dr.ir. J.M.H. VissersProf.dr. M. van Houdenhoven

Externe commissieProf.dr. C.J. KalkmanProf.dr. W.F. BuhreProf.dr. R.J. StolkerProf.dr. E.W. HansProf.dr. P.L. Meurs

274

MUMC

Cover - VU Research Portal

Documents